Opportunity

As of 2017, indoor rock climbing in the US was a $402.9MM industry by revenue, with an annual growth rate of 3.9%. This is an industry that sees growth with the growth of disposable income, and sees its growth driven primarily by younger consumers. Consumers age 35 and below comprise 83.6% of the market, with 47% of the market aged 25 to 34.

Despite this growth, safety concerns often dissuade new climbers. Additionally, high costs in the form of the climbing gym needing liability insurance and knowledgeable staff to train and keep an eye on the climbers, increases the barriers to entry in the market.

Further, to keep climbers continuously engaged, climbing gyms employ route setters – people who redesign the patterns of the climbing holds every few weeks. To communicate the level of difficulty of a route, routes are graded. While most indoor gyms are two standard grading scales – V Scale and Font Scale – grading is very subjective and there can be a high level of variance of how routes are graded across gyms. Further, individual attributes of the climber – such as height and weight distribution (men and women for instance, may have different weight distributions which will cause them to approach a climb differently) can make routes more difficult for some people than others. While there are six main forms of rock climbing holds, there can be a high level of variation on the main forms.

Solution

We believe an algorithm can be developed, factoring in types of holds, height and pitch of the climb, attributes of the climber, such as height and weight, to create new rock climbing routes with personalized ratings. Using artificial judgment in this way will create a more objective standard of route rating that will allow climbers to practice safely. Gyms across the country could then communicate a standardized definition of route difficulty, accessing a database of infinite climbing routes. We shall imbed the algorithm into a mobile app we named PocketClimber, both after the name of a rock climbing hold (Pocket) and the ease of using the app for setting new climbing routes.

Our solution reduces the time and cost for route setting. It will reduce the dead time between climbers as well as likelihood of injuries due to route-setting judgement error. Additionally, having routes analyzed by computers according to a person’s fitness and physical characteristics (such as height and weight) may make people feel more safe, thereby attracting new people to the sport.

Commercial value and promise

Our solution targets a rapidly growing international market and is scalable, as it can be used by individual climbers as well as large gym chains, such as Gold Gyms. PocketClimber can establish the standard for climbing and be adopted by all accredited gyms. We will design a prototype app to demonstrate the potential capabilities of the PocketClimber, pitching it to investors such as GoPro founder Nick Woodman who have a vested interest in adventure and sports. Live demonstrations will prove the viability of our product. Currently, no customized solution exists in the market. The closest is the app Vertical Life, which provides standard guides to climbing.

Pilot

We will start our pilot program with the demonstration of a prototype app to climbing gyms across Chicago and get feedback from climbers from our network and from the gyms. Additionally, we plan on hosting events around the app – such as human vs AI setters, and different climbing competitions sparking an interest in the app and in the concept. Climbers may be intrigued by the unusual routes designed by AI, so there could be an event-based opportunity here. After we collect enough data, both from the human setters and the climbers we will be able to refine our algorithm and demonstrate to investors the advantages of using PocketClimber with optimized routes and less accidents for beginner climbers.

Additional:

Features:

•      Wall scanner, some features of a climbing wall are not changeable. The app uses the phone camera to assess inherent difficulty of the wall
•      Hold Inventory:  Route setters can input the types of holds they have and how many holds
•      RouteBuilder: You can specify the level of difficulty you want, and RouteBuilder will make a route for you. Additionally, could provide further insight such as range of difficulty climbers will face, so that routesetters are taking diverse body types into account when setting  
•      RouteFinder: On the other side of the market, consumers could use the app as well to input personal metrics such as height in order to get personalized gym and route recommendations

Sources:

http://clients1.ibisworld.com.proxy.uchicago.edu/reports/us/industry/productsandmarkets.aspx?entid=4377#MM

https://www.99boulders.com/bouldering-grades  

Opportunity:

The global athletic footwear market size was valued at $64.3 billion in 2017 and the use of fitness apps has grown by 330% in the last 3 years. With fitness becoming an increasing part of a young professional’s life, finding ways to improve the efficiency of a workout or prevent injury are key to increased fitness success.

Solution:

Fun Run uses machine learning and sensors embedded in running shoes to track particular metrics and improve a runner/walker’s stride. In addition to tracking steps, heart rate (in the foot) and typical fitness tracking metrics that are currently available in the marketplace, Fun Run uses weight distribution to measure a user’s posture and stride. It also scans the running surface and analyzes data on both the surface type (soil, sand, concrete, etc.) and condition (dry, wet, icy, etc.). By continuously tracking and learning performance, the app can make recommendations based on a user’s goals. For example, if a user is concerned about hip pain, the app can let the user know if he or she is putting too much pressure on one side, then recommend relevant stretches/therapies and adjustments to the way they run in order to correct it. The algorithm can also use predictive analytics and alert people of the potential for injury or excessive soreness before people have experienced any pain.

Feasibility and Commercial Promise:

Athletes are increasingly interested in tracking their fitness. Of people who exercise at least monthly, 30% have a wearable fitness tracker and 29% use a mobile app to track fitness stats. Another ~25% plan to use these features in the future [1]. Despite significant commercial promise, the activewear market is saturating. 37% of fitness junkies state that brand is important, and 90% opt for high-end activewear brands [2]. As such, branding and high-performance will be integral to the success of Fun Run. Some of the features we plan to utilize may be expensive in the short term, such as sensors to analyze surface type and condition. The cost could be brought down by pulling in data from other sources (e.g., weather websites) or by relying on manual entry of some data.

Pilot:

A pilot would be rolled out at university track and field programs. We would work closely with a team of physicians and sports science staff at the University of Chicago (or comparable University), to certify and monitor all data and recommendations. We would seek to have each tailored recommendation and pain management solution be created and reviewed by our physician team to ensure proper legal protocol and medical viability.

Our pilot would seek out volunteer athletes who can provide large usage data sets (e.g., long distance runners) that would allow us to collect data on running strides, foot placement, weight pressure, shoe type and other relevant pieces of information. Additionally, we would bring the volunteers in for an initial screening process to create a baseline model of their postures, existing pain or medical issues, bone structure, etc. which would allow us to make better assessments from the running data we collect.

We would analyze the recommendations from our analytics software with our physician team to gauge relevance, accuracy, and benefit to the athlete above the normal utility of simple pain-based care.

Competitors/Risks:

A number of devices in the market measure biometrics such as heart rate and/or activity indicators such as steps and speed. They mainly allow people to track progress. Our product will be a leader in using AI to generate predictions based on such data combined with weather, running surface, and more granular biometrics information such as weight distribution. A competitive product on the market is Lifebeam VI, the self-proclaimed “first true AI personal trainer.” The product is a voice-activated Bluetooth biosensing headset with AI personal trainer. VI is focused on making you a proficient runner and is priced at $249.99. VI doesn’t analyze posture and weight distribution and doesn’t seem to be able to warn about potential injuries or excessive stress on your body. Other potential competitors are Google and Apple who can build on their personal assistants to also function as personal trainers. Existing user base and data will be a big advantage for them.

Sources:

[1] Mintel Report. Exercise Trends. US, October 2016.

[2] Mintel Report. Activewear. US, October 2016.

http://www.netimperative.com/2017/09/health-fitness-app-usage-grew-330-just-3-years/

https://www.engadget.com/2017/04/24/ai-personal-trainer-vi-headphones-running/

Team Members:

Sam Steiny

Rosie Newman

Gergana Kostadinova

Javier Rodriguez

 

Autonomous vehicle industry and the problem it solves

The global autonomous vehicle market is estimated to be $42B by 2025. This is segmented by Partially Autonomous Vehicles and Fully Autonomous Vehicles, representing $36B and $6B, respectively. The main benefit from autonomous vehicles is the expected increase in safety. Thousands of people die in car accidents and autonomous vehicles are expected to be less error-prone than humans. Autonomous vehicles also have additional benefits such as helping those with physical limitations to mobilize easier. Other benefits include: reducing the number of vehicles on the road, lowering the amount of traffic violations, and providing a more comfortable and efficient way of transportation. These solutions can then be expanded to optimize ride-sharing services and reduce traffic congestion and lower fuel consumption.

Optimus Ride is at the forefront of creating an entirely autonomous vehicle. It leverages a system of hardware products (and machine learning algorithms) in concert with human drivers to develop semi-autonomous vehicles. Initial use cases of the vehicles include shuttle services in communities, commercial developments, airports, college campuses, amusement parks, and other relatively low traffic areas.

The solution is two-fold – hardware augmenting people becomes software augmenting people.

Vehicles come with two lightweight Velodyne lidar sensors, eight cameras, GPUs and motion sensors, and a proprietary switchboard that translates the sensor data into mechanical responses from the vehicle. The system uses cameras and lidar sensors, but dependence on lidar will decrease as Optimus’ accumulates data, which will train computer vision algorithms. The system will eventually shift reliance to computer vision, which will use less expensive hardware and is more scalable.

Path to autonomous vehicle

Time to prepare, calibrate, test, and deploy a vehicle currently takes several weeks. That timeframe is expected to decrease as the company solidifies formal production partnerships with OEMs. Discussions are ongoing for several pilots, including some in Massachusetts and Florida. The product is currently comprised of a full-stack autonomous solution encompassing lidar-based perception (front and rear), vision-based perception (via front and rear cameras), motion planning (via wheel encoders), computer integration (via the NvidiaDrive PX platform), and drive-by-wire control. Lidar is the industry standard, but Optimus is working towards an advanced computer vision-based autonomous solution through a multi-layered rendering from three distinct visual input techniques: visual slam, deep learning, and stereo vision. The resultant camera-focused autonomous system, complete with sensor fusion, will represent the core IP and proprietary software. The unique solution will have the capability to detect objects and obstacles in the vehicle’s path and determine the location of the vehicle in proximity to its surroundings with centimeter-level accuracy. With this scalable turnkey solution available, Optimus could deploy full fleets of autonomous vehicles controlled by client platforms.

Roadmap to vision-based autonomy                Vehicle use cases

     

The fully autonomous vehicle landscape is highly competitive:

Robust proprietary data is a key competitive differentiator in the autonomous vehicle space. Driving data will train machine learning algorithms, underpinning self-driving technology. To scale, video or image-based data will be the most valuable because it can lessen dependence on expensive lidar technology by shifting reliance to computer vision and software. Incumbent players have accumulated road-mileage but capturing vision-based data remains to be an arms race.

Competitive Landscape

• Automakers (Ford, GM, Tesla) have been actively establishing partnerships with technology startups and making strategic acquisitions and investments in the autonomous vehicle space.
• Ride Hailing Companies (Uber, Lyft) are well-aware of the transition to self-driving cars and are developing in-house, through partnerships or via acquisitions.
• Technology Companies (Google, Apple)–Alphabet leverages GPS, Waze, and Google Maps to generate routes for autonomous vehicles and is developing autonomous offerings via Waymo.
• Autonomous Software Startups (Nexar, Nauto, Drive.ai, nuTonomy, Varden Labs, Aurora, NextEV)–have received significant funding to power the autonomous shift

But Optimus Ride benefits from a proprietary dataset and pilot partnerships:

Over time, Optimus Ride can leverage its multifunctional ride complex and license to operate in Boston’s self-driving vehicle zone, along with partnerships it has signed with private developers and community and transit authorities to use controlled city zones to accumulate driving data. One of the most obvious constraining market factors is the limited pool of talent, which is a crucial factor driving the numerous strategic acquisitions by incumbents that have defined the autonomous vehicle market in the last year. The technical expertise of the founding team provides a strong competitive advantage.

Team Members: Brentt Baltimore, Moises Numa, Corey Ritter, and Mitchell Stubbs

Sources:

https://techcrunch.com/2017/11/28/optimus-ride-will-provide-self-driving-vehicles-to-boston-community-residents/

https://www.bizjournals.com/boston/news/2017/11/28/self-driving-startup-to-offer-rides-in-weymouths.html

https://smartcitiesworld.net/transport/transport/optimus-ride-to-provide-autonomous-cars-to-modern-city-development

http://www.cbre.us/people-and-offices/affiliate-offices/new-england/new-england-media-center/optimus-ride-signs-new-seaport-headquarters0

https://www.bostonglobe.com/business/2017/06/12/officials-approve-another-self-driving-car-company-test-seaport/HEBBU5lURa6qh3LKHCP0pJ/story.html

https://www.optimusride.com/

http://www.thedrive.com/sheetmetal/15687/self-driving-car-startup-optimus-ride-gets-18-million-for-the-race-to-a-driverless-road

https://www.statista.com/statistics/428692/projected-size-of-global-autonomous-vehicle-market-by-vehicle-type/

http://loupventures.com/auto-outlook-2040-the-rise-of-fully-autonomous-vehicles/

 

Opportunity:

 

The Pennsylvania judicial system is one of many state prison systems that have been considering adopting statistically driven tools to determine how much prison time should be sentenced to individuals found guilty of committing crimes. The state spends $2 billion a year on its corrections system — more than 7 percent of the total state budget, up from less than 2 percent 30 years ago. Further, recidivism rates (the tendency of a convicted criminal to re-offend) remain high: 1 in 3 inmates is arrested again or re-incarcerated within a year of being released.  By properly identifying and distinguishing high / medium / low-risk offenders, the system has the opportunity to calibrate its sentencing accuracy to optimize the operations of its correctional facilities. In theory, risk assessment tools could lead to both less incarceration and less crime.

 

Solution:

 

The available risk assessment tools assign points to certain variables (such as age, gender, income, drug use, previous convictions, etc.) that have demonstrated to be strong indicators of criminal behavior in historical data. Social scientists have followed former prisoners and examined the facts of their life and monitored their lives for a number of years to develop an understanding of their propensity for repeated criminal activity. Many court systems use the tools to guide decisions about which prisoners to release on parole, for example, and risk assessments are becoming increasingly popular as a way to help set bail for inmates awaiting trial. This will ultimately help them save on their costs by providing better data-driven judgments towards criminal sentencing.

 

Commercial promise and challenges:

 

 

The main value proposition is that having an algorithm-based component to the judicial decision-making process helps many stakeholders through the value chain-

 

• Reduced risk of individual bias affecting judgments
• Increased efficiency reducing trial and bail time. Good for judges and defendants.
• Reduced costs which will inhibit better allocation of tax-payer money

 

While humans inherently rely on biased personal experience to guide their judgments, empirically grounded questions of predictive risk analysis play to the strengths of machine learning, automated reasoning and other forms of AI. One machine-learning policy simulation concluded that such programs could be used to cut crime up to 24.8 percent with no change in jailing rates or reduce jail populations by up to 42 percent with no increase in crime rates. Importantly, these gains can be made across the board, including for underrepresented groups like Hispanics and African-Americans.

 

On the other side, this approach faces challenges on an individual level especially because the system is based on a probability factored from similar offenders in the past that will influence the offender’s sentence despite that his future could be different or in other words an outlier to the statistics. To minimize these errors we will need to know whether the system will have enough variables to most accurately assess the individual as much as possible and whether these tools will supplement the judge’s decision rather than depend on it. 

 

Competition:

Even though a sizable amount of the agencies and organizations using AI systems in criminal justice reform are governmental bodies, the algorithms and software they use are privately owned. Due to the nascent nature of this industry, competition may either be among private companies trying to develop more efficient and fair algorithms, or there may be competition from an altogether different process, such as a community-based open-source AI project. A report from the Brookings Institute highlights the success of programs such as Google’s Tensorflow and Microsoft’s DMTK as proof.    

 

Proposed alteration:

 

The possible risk-assessment tools should be highly integrable with the existing software and processes in use in the justice system. While companies will want to utilize the ‘black box’ model that allows them to keep their algorithms confidential, it may lead to legal challenges such as in the ‘Loomis v. Wisconsin’ case (Wired). Thus, we would emphasize an open-source based solution with data security being prioritized.

 

Another difficult question in building the model is to tease out factors that are strong indicators in the prediction model without regressing to biases based on race and SES that are prevalent in the current judicial system and are socially deemed unfair.

 

Lastly, these tools could be enhanced by factoring in an inmate’s behavior in jail for their next trial to mitigate mistakes when they happen. If data shows that an inmate will likely not repeat a crime when they show good behavior during their sentence then this will provide us to have a further efficient system that will be as fair as possible while maintaining our goal of reducing costs on correction systems.

Sources:

 

https://www.themarshallproject.org/2015/08/04/the-new-science-of-sentencing#.wewudvPdi

 

https://www.brookings.edu/blog/techtank/2017/07/20/its-time-for-our-justice-system-to-embrace-artificial-intelligence/

 

https://www.wired.com/2017/04/courts-using-ai-sentence-criminals-must-stop-now/

 

https://www.thoughtworks.com/insights/blog/how-artificial-intelligence-transforming-criminal-justice-system

 

https://www.cs.cornell.edu/home/kleinber/w23180.pdf

 

https://www.uaa.alaska.edu/academics/college-of-health/departments/justice-center/alaska-justice-forum/34/3winter2018/a.pretrial-risk-assessment.cshtml

 

Team Members:

Mohammed Alrabiah

Tuneer De

Mikhail Uvarov

Colin Ambler

Lindsay Hanson

Opportunity

 

About 50% of the US adult population suffers from a physical condition, and a further 33% of those also have a mental illness. However, although there are people who suffer from both, it is often only the physical condition that gets treated. Studies have shown that people with physical health issues (heart failure, for example) and an untreated or undertreated behavioral health issue (such as depression or anxiety) cost 2-3x more for treatment of their physical conditions. Per 2012 data, those patients accounted for almost $300B annually in excess health care spend, mostly attributable to use of medical (as opposed to behavioral) services. Given that this is the total expected cost savings that a potential solution would provide to patients and insurance companies, we believe this is a reasonable estimate of the market size.

 

Solution

 

Quartet Health fixes this problem by bringing data-driven predictive analytics and recommendations to connect untreated or under-treated patients with physical ailments to the correct mental health providers, leading to a comprehensive treatment plan. Quartet accomplishes this by tapping into big data to identify people within a primary care system with undiagnosed or untreated mental health conditions. Quartet Health partners with insurance providers to analyze millions of insurance claims and flags patients with comorbidities or who have not been treated for behavioral health issues; it can combine results from behavioral health screenings with the patient’s data to reveal who might benefit from behavioral healthcare. It also matches those patients with local behavioral health providers who accept their insurance and who can meet both face to face and via telemedicine. Going forward, it then notifies the primary care doctor and follows up to make sure patients keep to their appointments, checks clinical results, and calculates the cost of care.

Effectiveness

 

Since hospital systems and insurers pay for Quartet’s platform, it’s free to use for patients as well as primary-care doctors and behavioral health specialists and hence using Quartet’s system lowers costs. Through partnerships with insurers and healthcare systems, health plans pay per member, and compensation is tied to quality of care and cost reduction. The shift from fee-for-service to value-based payment rewards providers for patient outcomes, compelling them to provide superior holistic treatment.

 

This also positively impacts patients’ lives because their treatment will be more comprehensive, effectively targeting both the physical and mental conditions, enabling them to live their lives more comfortably without need for return hospital visits and unnecessary expensive bills. The tool will allow doctors to proactively diagnose mental issues before they become exacerbated, leading to a) effective treatment of the mental condition and b) positive side effects in assisting the recovery of the source physical issue.

 

Competitive Landscape

 

Although the use of augmented judgment in diagnosing mental illnesses has a lot of potential, it is still in its early stages. For instance, NeuroLex is pioneering a computer model that predicts the onset of diseases, such as psychosis and schizophrenia, by collecting and analyzing patients’ speech samples for patterns (i.e. pauses in the words, use of determiners, etc.) that can be indicative of each disease. Other companies that operate in this space include New York-based AbleTo, which announced a $36.6 million raise for its behavioral health platform, U.K.-based Ieso Digital Health, which raised $24 million for a platform that offers psychological therapies and cognitive behavioral therapy, and Talkspace, which has raised $59 million to offer on-demand virtual therapy via instant messaging. Related to this are meditation platforms such as Headspace, which recently closed a $36.7 million round, while Y Combinator alum Simple Habit raised a small $2.5 million round to help the world destress.

 

To the best of our knowledge, however, the only other company taking a similar approach in using predictive analytics at scale to better align patient, provider, and insurance outcomes is Clover Health. We feel Quartet Health can compete with Clover because of its exclusive focus on behavioral health, as opposed to Clover, whose stated goal is overhauling the entire health insurance industry using big data.

 

Suggestions/Improvements

 

Quartet Health’s ability to create and extract value is tightly coupled to both the quality of its predictions and the quality of behavioral healthcare that patients receive. If they can improve their models’ precision without decreasing recall, they will identify more patients for whom they can improve care while reducing health providers’ costs. If they can improve health outcomes for identified patients, say by improving its matching to the right healthcare providers or by increasing patient compliance, this will further extend those benefits. To improve their predictions, Quartet Health should refine its algorithm by: (1) collecting more data on patients with confirmed physical and mental illnesses (so that causal patterns which would otherwise go unnoticed are identified) and (2) increase both the number of data points and types of data collected for each “potential” patient. Besides hospital records and current symptoms, studies have shown that an individual’s social media activity can predict suicidal intent or indicate mental illnesses. As such, sentiments, online behavior and notable changes in the way someone interacts with peers can be additional data points a patient can choose to provide to aid in diagnosis. The risk is that patients may feel their privacy has been compromised. To mitigate this, Quartet Health could first solicit doctor and patient approval, primarily monitor public activity, and only connect directly with the peers/friend-group of high-risk individuals.

 

Additionally, to improve patient outcomes, Quartet Health should incorporate more feedback loops so that it can provide recommendations based on what patients in the same age group and with similar conditions and medical records, found to be effective. Ways to do this include:

 

• App-Doctor: The patient can answer questions and post daily logs on mental progress, and the application can give further instructions on what to do based on these inputs, ideally forgoing the need for a doctor in less-critical cases.
• Patient-driven questionnaires: Have patients answer questionnaires when they are with their primary care doctors, and then use answers to immediately match patients to specific mental health professionals, drug treatments, etc. This has the added benefit of gathering data directly from patients.

 

Sources

https://www.quartethealth.com/

http://www.modernhealthcare.com/article/20171125/TRANSFORMATION03/171119895

https://www.forbes.com/sites/zinamoukheiber/2015/10/25/patrick-kennedy-backs-quartet-health-as-startups-in-mental-health-are-suddenly-hot/#127ea65b62fc

http://www.mobihealthnews.com/content/quartet-sutter-health-use-big-data-get-patients-mental-healthcare-they-need

https://www.quartethealth.com/blog/price-wrong-physical-costs-behavioral-health-issues

https://www.theatlantic.com/health/archive/2016/08/could-artificial-intelligence-improve-psychiatry/496964/

https://www.neurolex.ai/

https://www.linkedin.com/pulse/its-time-tech-put-human-touch-back-healthcare-arun-gupta/

http://www.healthcareitnews.com/news/sutter-health-quartet-health-partner-mental-health-coordination

https://venturebeat.com/2018/01/03/quartet-raises-40-million-to-bridge-the-physical-and-mental-health-care-divide/

https://www.healthcare-informatics.com/article/mobile/and-comers-2017-quartet-health-s-venture-behavioral-health

https://www.crunchbase.com/organization/talkspace

https://www.crunchbase.com/organization/clover-health

https://www.fastcompany.com/40513289/quartet-health-just-raised-40-million-to-expand-its-healthcare-platforms

https://techcrunch.com/2017/11/27/facebook-ai-suicide-prevention/

 

Team

Siddhant Dube

Eileen Feng

Nathan Stornetta

Tiffany Ho

Christina Xiong

 

• Opportunity

 

The recent increase in the number of satellites, providing multi-spectral, full-earth coverage at rapidly decreasing cost, provides significant opportunity to achieve business insights through the analysis of real-time or near real-time high-resolution satellite imagery. When coupled with significant increases in computing power and cloud computing networking, companies can use machine learning algorithms to interpret, analyze, and process imagery data to solve business problems across a large number of applications. At a simplistic level, “picture-takers,” or companies that operate a constellation of satellites, capture images in multiple spectrums (visible, infrared, radar, etc) of the earth at high resolution (down to 30 cm) for further processing. Next, a “sense-making” company accesses the data and analyzes it by comparing different images of the same object, aggregated across multiple objects and periods of time, in order to provide business insights for the given application. One of the leading companies in analyzing this satellite imagery is Orbital Insight.

 

 

• Summary of Solution

 

While some satellite companies also perform the subsequent analysis, Orbital Insight is purely a “sense-maker.”  They are experts in getting useful aggregate information out of the satellite images. In order to do this, they employ proprietary machine learning and computer vision algorithms to observe changes in specific activities or resources which are visible in or can be deduced from the satellite images.  The algorithms are trained to identify specific features in the images, like cars in a parking lot, crops in a field, or oil tankers at sea, and quantify them. In doing so, Orbital Insight can preempt commodity trends or estimate non-public information regarding a company’s performance. They partner with several key satellite constellation owners to gain access to higher quality, more complete, or more frequently updated satellite imagery.

 

 

• Evaluate Effectiveness / Commercial Premise

 

Orbital Insight solves business problems through both industry and product-specific channels. While Orbital Insight currently does the majority of its business through product-specific channels, analysis gleaned from satellite imagery can be applied to almost every industry where insight is possible through monitoring global trends or competitor performance. Among others, notable consumers of Orbital Insight’s analysis include retail, energy, financial services, agriculture, insurance, and government. By using satellite imagery, for example, Orbital Insight is able to forecast U.S. corn and soy production with predictive crop yield analytics through merging rich satellite data, weather and historical data in real time to provide investment-grade insights to a variety of clients including hedge funds, asset managers and financial data service providers before commercial and government statistics are available. In another application, Orbital Insight used satellite imaging in the aftermath of Hurricane Harvey to refine their model predicting flooding for their insurance company clients. Orbital Insight’s use of machine learning and data analytics applied to satellite imaging can provide customers with both valuable investment insights and business solutions.

 

• Competitive Landscape

 

As mentioned previously, the competitive landscape in the satellite imaging market is largely divided between “picture-takers” and “sense-makers,” and some companies try to span both business models. As a result, Orbital Insight competes against both imaging companies that perform analysis (e.g. Planet Labs) and companies with no satellites who focus solely on analysis (eg, Descartes Labs). While the satellite imaging industry (i.e. the “picture-takers”) is projected to be a $6.8B industry by 2023, the satellite imagery analysis industry in which Orbital Insight competes is less mature and composed largely of private companies, spanning industries that produce trillions of dollars in annual revenues. There are very few barriers to entry for imagery analysis, and thus Orbital Insights achieves its competitive advantage through partnerships with a number of the “picture-takers,” where the quality of analysis produced from its algorithms is higher due to Orbital Insight having access to more timely images sourced from different data streams (i.e. multi-spectral), and its leading proprietary machine learning algorithms. While Orbital Insight is a private company that was founded very recently, with a valuation of approximately $230M, DigitalGlobe, a leading public imaging and analysis company, provides an idea of the size of potential revenues, achieving $725M in revenue in 2016.

 

 

• Proposed Alterations to Increase Value

 

Orbital Insight has a strong business model with a clear value proposition, but operates in a competitive field with low barriers to entry.  They are entirely at the mercy of the “picture-takers” and are squeezed at both sides of the stack with little likelihood to succeed through vertical integration.  In order to succeed in the space, they should differentiate themselves on algorithm performance and take measures to protect their methods. If they are known as the player that can most accurately predict commodity trends and extract the most value from satellite images, then they will have a place in the ecosystem. Since the software patent landscape has varying levels of effectiveness based on the applicable country, is complicated, and in a state of flux, Orbital Insight should be more proactive than the average company in protecting its IP. Lastly, in order to not be beholden to any single upstream firm which may itself try to vertically integrate, they should also make efforts to expand the number of partners in the satellite imaging industry.  This will ensure that satellite imagery continues to be a widely available product, allowing companies like Orbital Insight to benefit from an ever-expanding amount of data.

 

Sources

Newsweek

http://www.newsweek.com/2016/09/16/why-satellite-imaging-next-big-thing-496443.html

AI Applications for Satellite Imagery and Satellite Data

https://www.techemergence.com/ai-applications-for-satellite-imagery-and-data/

How AI Could (Really) Enhance Images from Space

https://www.wired.com/story/how-ai-could-really-enhance-images-from-space/

Global Commercial Satellite Imaging Market Size, Share, Development, Growth and Demand Forecast to 2023 – Industry Insights by Application, and by End-User

https://www.researchandmarkets.com/research/dq6sc4/global_commercial

Orbital Insight Sees the Big Picture with AI

https://www.nanalyze.com/2017/01/orbital-insight-artificial-intelligence/

How Orbital Insight Measured Hurricane Harvey’s Flooding Through the Clouds

https://www.forbes.com/sites/alexknapp/2017/09/26/how-orbital-insight-measured-hurricane-harveys-flooding-through-the-clouds/#5a7fe27b676c

DigitalGlobe Form 10-K

https://www.sec.gov/Archives/edgar/data/1208208/000155837017001064/dgi-20161231x10k.htm

Patent Protection for Software-implemented Inventions

http://www.wipo.int/wipo_magazine/en/2017/01/article_0002.html

 

Team Members

Thomas DeSouza, Matthew Nadherny, Patrick Rice, Samuel Spletzer

Recursion Pharmaceuticals uses augmented perception to speed up the initial stages of drug discovery.

 The problem. Matching thousands of drugs to thousands of diseases:

There are thousands of rare diseases and thousands of FDA approved drugs that may have a positive impact on those diseases, but those drugs, both singly and in combination, on the rare diseases is time-consuming. Rare diseases are diseases that affect less than 200,000 people.  Pharma companies are less likely to pursue those smaller markets. Despite the small market of any single disease, it’s estimated that rare diseases affect 10% of Americans. If Recursion’s platform can help access this large market of 10% of Americans, it could be massively valuable.

Recursion was founded by an experiment to find a drug that impacts a rare disease which weak blood vessels lead blood into the brain and cause strokes. The experiment applied 2,000 different drugs to a diseased cell sample. Then, a pair of cell biologists looked at the 2,000 experiments to evaluated the phenotypic impact – or how the drugs appeared upon visual inspection to impact the cell sample. A computer algorithm developed by Anne Carpenter’s group at the Broad Institute also looked at the images from the experiments. The human team selected 39 drugs that appeared to have a positive impact on the diseased cells.

Here’s the interesting par: the computer program also selected 39 drugs – but the computer-selected and human-selected sets didn’t overlap at all. The computer selected 39 different drugs than the people did.

And, after closer study, only one of the people-selected drugs continued to appear to have an impact while 7 of the computer-selected drugs were impactful enough to merit further study.

The solution. Recursion’s automated phenotypic drug discovery platform:

Phenotypic drug discovery involves testing a drug in vitro (in a test tube) by applying a drug compound to a diseased cell sample. Researchers can judge the impact of the test by observing phenotypic changes in the cell sample.

Recursion’s platform has automated microscopes that sends thousands of images of in vitro tests each week to image recognition software. The software, armed with image data of healthy tissue, looks at the images and determines if the tested drug makes the cells look healthier.

Initial results. Recursion has made significant progress toward their goal of treating 100 genetic diseases by 2025:

Thus far, Recursion has identified promising compounds for 34 different rare diseases. Seven have progressed to in vivo tests and two are nearing applying to enter FDA trials. Their success has generated investor interest; Recursion has raised nearly $80M since February 2017.

 

 

Critique: the pharma value chain is challenging: 

The main issue that I can see with Recursion’s platform is that, because of the nature of the value chain in drug discovery, it fails to extract much value. Simply identifying that an approved drug has in vitro phenotypic impacts on a cell sample isn’t worth much. From there, millions of dollars must be spent to test, optimize, and characterize the lead compound in animal models. Then the drug must be tested in people in FDA clinical trials, which takes many years and many $Ms.

Recursion simply produces the early lead. Now, that drug is FDA-approved, so safety tests would often (but not always) less intensive. And, many pharma companies that own these drugs just have them sitting on the shelf, not being used or sold. So it could be a compelling tool for large pharma companies that have patented compounds that are failing to generate good data in the clinic – sort of a second-life option for those drugs.

 

But the dataset could be valuable in the long run:

However, Recursion also seems focused on generating a massive dataset of cellular models, which grows by 20 TB each week. This data could be a valuable tool for a large pharma company that develops its own drugs, such as Recursion partner, Sanofi. I would suggest focusing on collecting that data. Perhaps the company could develop a portable version of their platform and partner with CROs (contract research organizations) to collect data from clinical trials.

 

However, in the long run, competition from AI drug design companies may be problematic:

One general category of competitors are general AI-enabled drug development companies, such as Insilico Medicine and Atomwise. These companies promise to perform rational design of drugs from computer models. However, they haven’t yet produced compelling results.

 

Sources

Recursion: https://www.recursionpharma.com/

NIH profile: https://sbir.nih.gov/statistics/success-stories/recursion

NIH rare diseases: https://rarediseases.info.nih.gov/diseases

Anne Carpenter’s group at the Broad: https://personal.broadinstitute.org/anne/

Investor blog post about Recursion: https://medium.com/@CRVVC/recursion-pharmaceuticals-ai-enhanced-drug-discovery-fdb8d7aad64c

Tech Crunch article about $60M Series B: https://techcrunch.com/2017/10/03/drug-discovery-startup-recursion-raises-60-million-in-series-b-from-dcvc/

FierceBiotech profile: https://www.fiercebiotech.com/special-report/recursion-pharmaceuticals

 

 

Team Members: Brentt Baltimore, Moises Numa, Corey Ritter, Mitchell Stubbs

Opportunity:

The US Luxury goods market is $85B and growing slowly, worldwide the market was $249B in 2016. The luxury goods industry has seen a recent foray into augmented intelligence with the advent of the Apple watch. At the Basel Fair in March, de Grisogono, one of the biggest jewelers in the world launched a product-driven chatbot, that guides users into selecting various types of jewelry (rings and pendants in this case). The chatbot first introduces itself and then compliments the customer, finally proceeding to ask questions about the customer’s taste and using augmented intelligence eventually offering a choice of jewelry to buy.  The industry in general is facing a decline with big players like Tiffany & Co. seeing declining sales and profits in the last two years. As Millennial tastes move away from jewelry and traditional diamonds, companies like De Beers are focusing on ways to improve customer experience. This brings us to our solutions to improve this market.

We believe a potential enhancement to this industry would be an app to virtually try on jewelry. Data-driven marketing utilizing customer CRM data and on-site browsing habits would ensure customers tastes are met in the best manner. WIth the help of the AI demonstration and experience, retailers such as Tiffany & Co could potentially decrease the size of their brick and mortar stores, reducing rent costs.

Effectiveness and commercial promise:

This strategy seems like it will pay off in a large way. According to statistics by McKinsey, online sales of luxury goods have been increasing relative to overall sales, showing an increasing willingness to make large purchases virtually. By creating augmented displays, these companies decrease the time and effort required for a consumer to “try on” their product and increase the variety of available options to test out with smaller operational costs. These tools could be easily scaled with little additional marginal cost.

One limitation to this effort’s success is the ease of reproducibility. Because other competitors in the space can easily copy any successful initiatives, it may not serve as a strong differentiator for any one firm. However, it should drive further sales industry wide by lowering the cost of creating an endowment effect by showing the potential customer what it would look like if they wore this particular jewelry.

As the chatbot and the photo get information on customer preference, the company can further personalize the offerings. In this case, the phone acts as a sensor, returning information on how long users interact with the app and how many options they are choosing between. Information from photos that users take with the app would sense information as well. For example, the app could suggest different jewelry for different outfits. Chatbots are an easy win for luxury retail. The more data is gathered by the chatbot, the more personalized the price can become too, allowing for capture of the maximum consumer surplus.

Chatbots are replacing human customer service through online chat. The more bot use, lead to decreased cost and higher service value for the customer. For a concierge, the efforts to ask many questions, the time it takes to analyze it and the potential for error in recommending a new product are all costly. Whereas once a chatbot is programmed well, the cost is very low and the service is available 24/7/365.

The chatbot  will assist in gathering consumer preferences and then the virtual “fitting room” will be used to determine whether the customer is satisfied with the item. There is significant effort in the online retail space to develop the technology needed for an accurate virtual fitting room. If the experience is not accurate, it could adversely affect the credibility of the retailer and its technology. Amazon has introduced the Echo Look style assistant, which has received positive reviews.

The virtual stylist is being applied to fashion, there is an opportunity to bring it specifically to the jewelry space. The company GRANI is an early adopter of the virtual jewelry fitting room space. The design will be refined, allowing users to try jewelry with different outfits/ hairstyles, improving the image quality so the exact cut and quality of the jewelry is apparent.

LINKS:

https://www.wsj.com/articles/tiffany-hunts-for-path-to-regain-cool-1499621248

https://thinkmobiles.com/blog/augmented-reality-jewelry/

https://www.fool.com/investing/2018/03/17/why-tiffany-co-stock-dropped-on-friday.aspx

https://www.mckinsey.com/industries/retail/our-insights/luxury-shopping-in-the-digital-age

https://www.ft.com/content/1c2a6b24-a514-11e7-8d56-98a09be71849

Market size: https://www.luxurysociety.com/en/articles/2017/07/us-luxury-goods-market-sees-another-year-slow-growth/

http://www.bain.com/publications/articles/luxury-goods-worldwide-market-study-fall-winter-2016.aspx

Personalized Pricing http://review.chicagobooth.edu/marketing/2018/article/are-you-ready-personalized-pricing

Analogy to Chatbots: https://chatbotsmagazine.com/3-high-value-chatbots-types-and-1-you-need-to-fire-immediately-49832901fe8a

https://www.econsultancy.com/blog/66058-fashion-ecommerce-are-virtual-fitting-rooms-the-silver-bullet

https://www.prnewswire.com/news-releases/facecake-releases-first-online-mobile-and-in-store-augmented-reality-shopping-platform-for-jewelry-at-nrf-2018-300583203.html

https://www.retaildetail.eu/en/news/mode/amazon-brings-virtual-fitting-rooms-your-home

https://www.trendhunter.com/trends/try-on-jewelry-pieces

Team Members:

Jess Goldberg

Anu Mohan

Louis Ernst

Pranav Himatsingka

Andrew Herrera

The NextGen Solution to Your Perfect Photo Needs

The Problem / Opportunity

The US photography market has $10 billion in annual revenue. A number of startups and large technology companies have aimed to improve both professionals and amateurs’ photographs (especially those taken for social media), principally by focusing on post-production services, achieving valuations in the hundreds of millions. To our knowledge no technology company has focused on the pre-production photography component, leaving significant  open space for our company.

According to a recent National Geographic’s 50 Greatest Pictures issue, “a photographer shoots 20,000 to 60,000 images on assignment. Of those, perhaps a dozen will see the published light of day”. Photography is an art that depends on a number of factors – timing, weather, sun exposure, angle, and more – all of which lend to the unfortunately ephemeral nature of the perfect snapshot. This problem creates a great opportunity for a tool that can decrease the amount of time, energy, and planning needed to capture the optimal image.

 

Solution & Data Strategy

Focus Pocus creates a solution that can track, identify, and predict the best locations for a photo, enabling casual and professional photographers to see where they should navigate to in order to capture their ideal shot. This solution will be made available as a downloadable app on the user’s phone. In future iterations, Focus Pocus may be installed natively into wifi-enabled cameras.

Focus Pocus solves the problem of finding and taking the ideal shot by crowdsourcing the best possible photograph locations and conditions, relying on large amounts of publicly available data combined with sensor data (from users’ cameras/input) and guiding the user through the photo setup process.  

First, Focus Pocus will integrate with photo-sharing platforms like Instagram, Flickr, Google Photos, and 500px to identify publicly available photographs that are either (1) highly popular or (2) of a high quality for the area you are located. Highly popular photographs on social media can be measured by how frequently each photo is clicked, shared, or liked. High quality photos can be identified using deep learning photo-scoring algorithms that can identify good photographs based on characteristics such as clarity, uniqueness, color, etc.

Next, Focus Pocus will identify which ideal shots are available to you and what settings or angles you need to use in order to achieve them, based on your camera type, time of day, lighting conditions, and other user-specific data.

Most of this data is available; photographs taken using non-phone cameras will typically contain large amounts of technical metadata under the Exchangeable Image File Format which includes the following:

• Date and time taken
• Image name, size, and resolution
• Camera name, aperture, exposure time, focal length, and ISO
• Location data, lat/long, weather conditions, and map

Over time, after initial training and being provided with photo datasets, Focus Pocus can continue to map out the entire city, with the goal of handling everything from recommending tourists photo spots to internally setting up the camera with the right specs and using AR to position the camera at the right height and distance from the subject. Tracking sunlight and weather conditions by utilizing training data to identify the best locations using current time and conditions can also be developed (by using integrations like LinkedIn and Rapportive).

 

Pilot & Prototype

The project lends itself to piloting at trivial cost in a single city before scaling to other cities. As a pilot, we would set up sensors (light and weather sensors, cameras, etc.) at a handful of highly-trafficked photo locations in Chicago, determined by assessing geospatial photo density using a service like TwiMap or InstMap. Early candidates would be the Bean, Navy Pier, Millennium Park, and Willis Tower. Focusing on these locations initially would also make it easier to market the product with concentrated advertising or founding employees giving demonstrations on-site.

From there we would develop a simple mobile photography app for users that would be used to recommend ideal photo locations and also enforce ideal camera settings within the app for a location based not only on the phone’s sensors, but also from our more refined on-site sensors.

 

Validation

To ensure that our solution meets the objectives of identifying the best locations and conditions for photographs, we could evaluate the predictive power of sensors by benchmarking photographs taken using Focus Pocus against those without. We could also measure the following success metrics:

• Number of Downloads/Installations, Retention Rates, Usage Per Member

In validating market need, we’ve also done research on applications similar to Focus Pocus and have found that businesses, such as Yelp and Flickr have already used deep learning to build photo-scoring models. For instance, by assessing factors, such as depth of field, focus and alignment, Yelp is able to select the best photos for its partner restaurants. However, this use case is after-the-fact (i.e. after the photo has already been taken), whereas the solution we propose allows for actions to be taken to optimize photo quality before it’s taken.

 

Sources

https://engineeringblog.yelp.com/2016/11/finding-beautiful-yelp-photos-using-deep-learning.html

https://digital-photography-school.com/1000-shots-a-day-the-national-geographic-photographer/

Jin, Xin, et al. “Deep image aesthetics classification using inception modules and fine-tuning connected layer.” Wireless Communications & Signal Processing (WCSP), 2016 8th International Conference on. IEEE, 2016.

Aiello, Luca Maria, Rossano Schifanella, Miriam Redi, Stacey Svetlichnaya, Frank Liu, and Simon Osindero. “Beautiful and damned. Combined effect of content quality and social ties on user engagement.” IEEE Transactions on Knowledge and Data Engineering 29, no. 12 (2017): 2682-2695.

Datta, Ritendra, and James Z. Wang. “ACQUINE: aesthetic quality inference engine-real-time automatic rating of photo aesthetics.” In Proceedings of the international conference on Multimedia information retrieval, pp. 421-424. ACM, 2010.

https://www.crunchbase.com/organization/magisto

https://www.crunchbase.com/organization/animoto

https://hackernoon.com/hacking-a-25-iot-camera-to-do-more-than-its-worth-41a8d4dc805c

https://twimap.com/

https://instmap.com

 

Team Members

Siddhant Dube

Eileen Feng

Nathan Stornetta

Tiffany Ho

Christina Xiong

Opportunity:

Descartes Labs is a startup founded in New Mexico in 2014 that is building data refinery for satellite imagery to better understand the planet. Currently, even with so much data and imaging of the plant available, it is difficult for companies and government agencies to successfully predict crop yields and potential shortages. This poses a challenge when preparing in advance for the changes year over year and increases the concern for climate change and food scarcity around the globe. Descartes claims that it can accurately predict crop yields, beating out the accuracy of the US Department of Agriculture, which is currently the only alternative for information.  

Solution:

Descartes uses the increase in the availability of large data sets accumulated by the increase in shrinking and cheaper sensors, as well as the rise in popularity of nanosatellites to determine how healthy the corn crop is on the planet from space. The company uses spectral information (non visible to the human eye) to measure chlorophyll to make these predictions and analyzes satellite data of every single farm in the US on a daily basis to update its predictions and deliver local estimates. 

Effectiveness, Commercial Promise, and Competition:

In terms of effectiveness, Descartes Labs states that it “can predict the yield of America’s 3 million square kilometers of cornfields with 99% accuracy.” Additionally, in 2015, the predictions made by Descartes beat those of the United States Department of Agriculture by 1% and the algorithms of the company continue to improve year over year.

Descartes Labs presents an opportunity for a wide range of groups, including corporations, government leaders, and humanitarian groups. For example, Cargill, an agricultural conglomerate, is a customer of and investor in Descartes Labs. The technology likely helps Cargill understand crop yields for a given year. Descartes Labs also received a grant of $1.5 million from the U.S. Defense Advanced Research Projects Agency, which uses the technology to anticipate food shortages, and thereby predict areas of sociopolitical conflict, in the Middle East and North Africa.  

Another application of the technology is disease forecasting and prevention. The high resolution private and public satellite data can help identify high risk environments such as areas with stagnant water conducive to mosquito proliferation. Those leads can be combined with medical and social media data to predict and backtest the spread of diseases. Such information will be valuable for epidemiologists and local governments.  

Several competitors include Orbital Insights, Gro Intelligence, and Tellus Labs. Orbital Insights covers a much wider range of industries – for example, it can help retail companies understand vehicle counts and traffic monitoring.

Suggestions / Improvements:

To improve, Descartes could utilize it current data and connect with various sensors on other devices, to triangulate the information it has and make more accurate predictions. This is a direction that CEO and Co-Founder, Mark Johnson, wants to go, given the vast amount of “potential sensor data we’ll be getting from combines, tractors, cars, boats, barges, trains, ships, grain silo. Everything is going to have sensors on it, so making sense of all that data is the sort of challenge we’re aiming toward” (Mark Johnson in Fast Company).

Descartes can also explore ways in which their data and capabilities can benefit individual farmers in addition to commercial clients such as Cargill. Descartes can partner with NGOs, consultants, and local governments to enable subsistence farmers with its data and technology.  

Another potential application is to tackle wildfires. Descartes can combine weather, geo imaging, and historical data from previous wildfires to identify high risk areas and potentially suggest effective ways for wildfire suppression once fires break out.

Sources:

https://www.descarteslabs.com/

https://www.fastcompany.com/40406046/this-startup-is-building-a-fitness-tracker-for-the-planet

https://medium.com/@thephilboyer/announcing-our-investment-in-descartes-labs-9dca8257d0d9

https://www.forbes.com/sites/themixingbowl/2017/09/05/can-artificial-intelligence-help-feed-the-world/#110f4bd346db

https://blog.nationalgeographic.org/2018/02/21/forecasting-diseases-one-image-at-a-time/

https://venturebeat.com/2017/08/24/descartes-labs-raises-30-million-to-better-understand-earth-with-ai/

https://www.theverge.com/2016/8/4/12369494/descartes-artificial-intelligence-crop-predictions-usda

Clients can integrate their data with the Descartes Platform to create their own solutions, models, and forecasts:

Team Members:

 

Sam Steiny

Rosie Newman

Gergana Kostadinova

Javier Rodriguez