AiThority Interview with Priya Ponnapalli, Senior Manager at Amazon Machine Learning (ML) Solutions Lab

Hi Priya, please tell us about your role at AWS and how you arrived here.

I lead the Amazon Machine Learning Solutions Lab (MLSL), where my global team of scientists, engineers, and product managers help AWS customers identify and implement their most important ML opportunities. 

I enjoy working at the intersection of multiple disciplines and applying science to solve real-world problems. Throughout my career, I’ve built and delivered high-impact ML products to solve key challenges in all industries. The opportunity to work with diverse customers across a breadth of industries and business problems is what attracted me to the ML Solutions Lab. I’m passionate about making ML accessible to all, and I get to work directly with customers to help them innovate using ML.

Tell us about your biggest milestones from your background prior to AWS?

I’ve always loved math, and my passion for the subject continued throughout university and made me want to pursue a career in science. 

Coming out of undergrad, I was very interested in digital signal processing. A class in graduate school at the University of Texas at Austin on signal processing for genomics turned out to be pivotal for my career, as it was my introduction to machine learning, and led to one of my biggest milestones; the professor of the class, Orly Alter, invited me to join her lab and she became my PhD thesis advisor.

Building on that foundation and my thesis, Alter and I co-founded Eigengene, which uses artificial intelligence to analyze cancer genome data and create personalized diagnostics and prognostics. We developed algorithms designed to find patterns within diverse, high-dimensional datasets known as tensors.

As I conducted research related to tensor decompositions, I realized that these algorithms are data- and industry-agnostic, with broad applications within multiple areas. This only furthered my passion for working across industries.

Recommended: AiThority Interview with Jordan Suchow, Assistant Professor at Stevens Institute of Technology

How did you cope with the pandemic months? What helped you stay focused and motivated?

At the start of the pandemic, like most people, I was very stressed and stopped taking care of my health. What helped me cope was exercise, a healthy reading habit, and taking on meaningful work such as COVID-19 safety efforts on behalf of Amazon and our customers. The pandemic has taken a toll on everyone in different ways, but the data shows moms are particularly affected. Amazon was very supportive in giving employees flexibility in the ways that we worked to be able to fulfill all of our many new obligations. 

The pandemic also significantly impacted the sports landscape as live sporting events ground to a halt and the industry grappled with ongoing restrictions as the world opened back up. If anything, it made my team more determined than ever to listen to our customers and work backwards from their needs. We helped Formula 1 (F1) transform its business of delivering content to broadcasters through remote broadcast production.

For background, the conventional way to televise a live racing event was to employ a travelling crew of 200-300 staff to attend all 23 races on the calendar, each of which requires over 1,000 4K streams and needs to be broadcast live back to F1 Headquarters in England. F1 was already transitioning to a remote production workflow prior to the global pandemic, but COVID-19 accelerated these efforts and AWS provided the infrastructure and speed they needed to pull off remote production in just seven weeks. AWS also helped power the 2020 NFL Draft, as they had to shift to a remote event due to the restrictions around large gatherings. This meant that players would not be able to gather in the green rooms and teams wouldn’t have representatives on the main floor. AWS supported the remote event to ensure that over 100 live feeds were running successfully, creating a seamless experience for the NFL, teams, coaches, players, fans, and everyone watching. It was the NFL’s first ever remote draft and the most watched ever, reaching more than 55 million viewers total.

A lesson you would like to give to every professional looking to join the AI ML domain-

Machine learning is now a core part of so many products and businesses, and we want it to work for everyone. Our diverse perspectives help us push each other to think bigger, and differently, about the products we build.

We all have an important role to play in shaping products of the future. With so many opportunities in machine learning, from scientists to product managers to engineers, anyone with an interest should not be afraid to pursue a job in the field. Given the availability of online resources, it’s never been easier to learn. In fact, as part of Amazon’s commitment to help 29 million people around the world grow their tech skills with free cloud computing skills training by 2025, we offer many opportunities for those looking to join the AI/ML domain, such as AWS re/Start, AWS Educate, AWS Cloud Quest, and AWS Skill Builder.

What are the unique projects and researches you have undertaken at the lab? 

The NFL Digital Athlete: The NFL and AWS are using data and analytics to co-develop the Digital Athlete, a virtual representation of an NFL player. What makes this so innovative is that the Digital Athlete gives us the ability to test out new safety equipment, rules changes, and allows us to predict injury events and recovery time for players. This Digital Athlete has been developed using machine learning techniques coupled with sensors and other data gathering tools to make it an exact, digital replica of an NFL player. The player can be adjusted based on playing position – depending on if we want to look at a quarterback, running back, defensive lineman, etc. – and can also be put into a near-infinite number of game scenarios, or environmental situations, to see how it performs. The Digital Athlete understands the rules of the game and how to play the game and can catch the ball, go into contact, tackle another player, or get injured – just like a real athlete. Using the Digital Athlete, the NFL can identify when injuries occur, and identify trends, patterns, and the forces involved with the injury. For instance, through data, the NFL saw that there was a trend around injuries during punts and kick-offs because of the speed in which the players would collide with each other. So, using this data, the NFL shortened that distance to reduce the number of injuries to players. We also launched an artificial intelligence challenge with the NFL to help improve our work on the Digital Athlete. This challenge crowdsourced ways for computers to automatically identify players using NFL game footage. Computer vision models created through the challenge achieved the goal of reaching human equivalence, meaning that computers can now automatically identify players involved in impacts from game footage with greater accuracy and 83 times faster than a human conducting the analysis manually. New models achieved a 40% increase in accuracy over existing models and represent a significant breakthrough for the NFL’s injury prevention and prediction efforts.

Swimming Australia: Swimming Australia reached out to AWS almost two years ago to help them unlock the power of data. Through an internal workshop Swimming Australia held with AWS Machine-Learning labs in the United States, they discovered they had many systems that were collecting different types of data. Some systems captured local events and competition results, while the others captured training camps, physiology, and medical data. While these systems were built for a specific purpose and do a great job for that purpose, Swimming Australia wanted to be better by being faster at getting their data and turning around insights.

As a result, Swimming Australia’s data was integrated into a data lake, which can transform data from different source systems that were in different formats into a unified data platform that can be queried efficiently from many services and interactive dashboards. For example, in the lead up to the Tokyo Olympics, Swimming Australia built an automated tool that can be used to support a coach’s decision-making in competition to predict possible and probable relay teams in the competition and tactical racing strategies. Previously, relay data was collected manually in excel, which meant that coaches were not able to compare performances beyond what was provided in the available static reports. Now, Swimming Australia has one report that automatically refreshes every day with data integrated from multiple sources, such as Omega, which is their timing equipment, and SwimCentral API. This process saved analysts a week on average compared to manual data comparisons. The best news of all was that accuracy was shaved to 0.1 seconds of what was forecasted. With this new capability, Swimming Australia was able to build performance comparisons between local and international competitions and started comparing the world’s best with the elite athletes in the organization to see how they compare at every stage of the race. Swimming Australia has a report for athlete ranking that ranks the local athletes across all the race events captured by different systems. Having all this data in one place allows coaches to see an athlete’s performance across events, including understanding an athlete’s development progression and national standardized testing protocols. This is especially helpful when it comes to relay races, as Swimming Australia can pull larger amounts of data quickly and predict potential relay orders for other teams. The team brought home nine gold medals in Tokyo. 

Recommended: AiThority Interview with Tim Harsch, CEO at Owler

What role does the ML Solutions Lab play at AWS? What role does it play particular to sports?

Through discovery workshops and ideation sessions, the ML Solutions Lab works backward from a customer’s business problem to deliver a roadmap of prioritized ML use cases with an implementation plan to address them. We then go step-by-step through the process of developing a machine learning-based solution for a customer’s organization. We also offer training to augment the level of ML expertise on a customer’s team including developer training, business leadership training, and a hands-on event through the ML Embark Program.

The ML Solutions Lab brings to every customer engagement the learnings from more than 20 years of Amazon’s ML innovations in areas such as fulfillment and logistics, personalization and recommendations, computer vision and translation, fraud prevention, forecasting, and supply chain optimization.

The ML Solutions Lab has extensive experience working with sports organizations to improve the fan experience and quality of the game. For instance, in AWS’s partnership with the NFL, the ML Solutions Lab identified and built an entirely new way for fans to engage with the sport through Next Gen Stats. Since 2017, AWS has been the NFL’s official technology provider in every phase of the development and deployment of Next Gen Stats. AWS stores the huge amount of data generated by tracking every player on every play in every NFL game — nearly 300 million data points per season. Amazon SageMaker is leveraged to quickly build, train, and deploy the machine learning models behind their most sophisticated stats, and the NFL uses the business intelligence tool Amazon QuickSight to analyze and visualize the resulting statistical data.

What are the sports technologies that most excite you?

In Sports, as in every industry, data, machine learning, and high-performance computing (HPC) are ushering in the next wave of technical sports innovation. Sports serve as a great medium to bring the benefits of technology to life. 

AWS works with some of the largest sports organizations in the world, and with each of these customers, we work backward to help solve their business problems through innovation. One of the most exciting examples that come to mind is my team’s work to prepare Swimming Australia for the Olympics by using machine learning to fine-tune the members and orders of relays. The result was, medals in 6 out of 7 relays in the 2020 Tokyo Olympics!

Could you tell us how AI ML is transforming the global motorsports performance and fan experience? We would love to hear about your findings related to Formula1 and NASCAR- please expedite with your insights. 

AWS is partnering with F1, Ferrari, and NASCAR to fuel digital transformation and unlock features to better educate, engage, and entertain fans in each sport. 

F1

• F1’s data scientists are training deep-learning models with 65 years of historical race data to extract critical race performance statistics, make race predictions, and give fans insight into the split-second decisions and strategies adopted by teams and drivers.
• Recently, F1 took the competition to the next level by altering some of the rules around car design for the 2022 season. Using The Computational Fluid Dynamics (CFD) project, F1 tests the aerodynamics of cars while racing, carrying out detailed simulations using AWS HPC services to make sim cycles faster and more sophisticated. By using AWS, F1 was able to reduce the average simulation time by over 70% —from 60 hours down to 18. F1 has been able to specify a car with only 10% downforce loss at the same 1 car length distance.

Ferrari

• Ferrari leveraged AWS using HPC to run thousands of simulations to gain insights faster than running simulations, encouraging experimentation and innovation in design.
• Scuderia Ferrari is using AWS technologies to build a new digital fan engagement platform via a mobile app. The app will deliver a fully personalized experience that informs, educates, and entertains fans about one of the most storied teams in racing history.

NASCAR

• AWS Media Services help NASCAR provide over 80 million fans instant access to the driver’s view of the race track during races, augmented by audio and a continually updated leaderboard. To ensure the highest-quality video stream and lowest latency for fans who choose to experience race day on a second screen, NASCAR Digital uses these tools to package, process, and store the video for delivery via Amazon CloudFront. 

What do you find most fascinating about leveraging machine learning in sports?

Of the various projects that I work on at AWS, I find the ML solutions applied to sports most fascinating. 

One of the best examples is our work with NFL on the Digital Athlete to prevent head injuries. Using the technology that developed the Digital Athlete, the NFL and AWS were able to test helmet performance to see how effective they were in preventing head injuries. As a result, the NFL shared this analysis and insight with players, and it led to a behavioral change. Prior to this research, only one-third of players were wearing top-performing helmets, but after this research, nearly 100% of players moved to the higher-performing helmets. This led to a 24% drop in head injuries in the 2018 season, and the same again in the 2019 and 2020 seasons. We have also developed helmet tracking capabilities, which can track every helmet on the field using computer vision.

By understanding the explicit forces causing injuries, including repetitive injuries, The NFL can, for example, change equipment or playing conditions to better protect players. This is also leading to the ability to identify safety rules changes and simulate the effects of the changes before they are implemented, ensuring that they have the desired in-game outcomes. The Digital Athlete has the potential to revolutionize injury detection and prevention in the NFL and many other sports beyond.

Think of the potential of this technology for sports like rugby, hockey, or any other contact or team sport.

Although the effort is focused initially on reducing the risk of head injuries, we are also expanding the technology to focus on other parts of the body, such as reducing the risk of lower extremity injuries, such as foot, ankle, and knee injuries, and measuring the impact of rules changes on these parts of the body.

Recommended: AiThority Interview with Adrian Jones, Chief Revenue Officer at ABBYY

How are different sports teams and leagues leveraging machine learning?

Sports partners use AWS to build cloud-based video archives that will automatically tag specific frames, from hours of video, with metadata so that they can easily search historical footage and surface pivotal moments. In tandem, real-time data crunching, analytics, and machine learning help enrich broadcast commentary and provide predictive insights. This enhances and enriches the fan experience before, during, and after games through rich data, insights, and second-screen experiences. 

For instance, Bundesliga is providing advanced real-time statistics and in-depth insights called Bundesliga Match Facts and will bring personalized experiences to its over 500 million fans, who gain more advanced insights into players, teams, and the league because AWS’s leading technology.

The NHL also applies AWS’s deep portfolio of machine learning services to game video and official NHL data – including data from the NHL’s Puck and Player Tracking (PPT) System and from the NHL’s Hockey Information & Tracking System (HITS) real-time stats – to develop and share advanced game analytics and metrics that take fans deeper into the game. 

Apart from sports, which other domains are you targeting to grow your AI ML opportunities? 

The ML Solutions Lab has successfully assisted customers from around the world across a diverse spectrum of industries including manufacturing, healthcare and life sciences, financial services, sports, public sector, and automotive to create new machine learning-powered solutions. A few examples include:

• We help manufacturing companies improve their core production, maintenance, safety, and quality, as well as R&D and supply chain functions. 
• We work with automotive customers in areas including supply chain optimization, in-car entertainment experiences, and autonomous driving. This includes improving the safety of drivers and pedestrians with accurate road scene perception and advanced analytics. 
• We work with financial institutions such as banks, investment organizations, insurance companies, and mortgage firms to improve forecasting, enable surveillance systems to flag new or emerging threats, generate personalized recommendations for financial products, automate document processing, and improve the customer experience with chatbots and conversational interfaces.

How should companies approach and build their data strategy? What kind of preparedness should a company have before leaping into AI Machine learning?

We are in the golden age of machine learning. We see businesses going from piloting ML projects to it having an impact on production. Many of the ML use cases currently in production are helping companies stay competent and resilient. It’s no longer a question of whether you should have an ML strategy, but rather how quickly you can put an ML strategy to action.

There are four important steps companies can take when building their data strategy:

1. Get the data in order.
2. Identify the right ML use cases that deliver the most value for the business.
3. Develop the culture of ML in the organization — this could be through upskilling workforce and educating both technical and business leaders.
4. Embrace the culture of ML. Iterating on things, embrace failures, and repeat to see that transformation.
5. We are seeing a dramatic shift among fans toward live TV streaming, real time fan engagement and delivering content on AR VR and metaverse. What is your take on leveraging AI ML to improve overall fan experience, and at the same time, make it more transparent for every audience? 

The fan experience is changing, and fans are getting deeper insights through visually compelling on-screen graphics and interactive second-screen experiences powered by AWS. For example, Second Spectrum uses AWS to minimize latency in its streaming services and offers fans a high-quality viewing experience. Rich data and insights reveal the nuances of in-game decision-making and highlight performances through advanced stats. 

Additionally, Pro Football Focus (PFF) leverages millions of data points from football games to uncover never-before-seen metrics that change how everyone experiences the sport. Automated data collection from sensors and cameras deliver real-time stats, guide game-time decisions, and help teams discover new ways to connect with fans.

Your predictions on the future of AI ML and data science in sports and digital content:

I’m excited to see how sports injury and prevention progresses. Through the use of rich data, AWS helps NFL teams innovate around training, health, and safety to further develop key components of the game. The virtual representation of a composite NFL player will enable us to eventually predict injury and recovery trajectories. The Digital Athlete has the potential to revolutionize injury detection and prevention in the NFL and beyond to other industries.

Thank you, Priya! That was fun and we hope to see you back on AiThority.com soon.

[To share your insights with us, please write to sghosh@martechseries.com]