Get to know more about our AI Sales Director Jit Seah Sam Au, his passion for AI, and his journey so far with the AWS DeepRacer League.
As the Formula 1 Singapore Grand Prix kicked into high gear, another race inches closer to its exciting conclusion—the AWS DeepRacer League, the world’s first global autonomous racing league.
Programming meets the pavement as developers worldwide come together to flex their machine learning (ML) prowess through a racing competition. Participants build reinforcement learning (RL) models to run on an AWS DeepRacer device and/or in a 3D virtual racing simulator.
Whether one is a seasoned pro or developing a model for the first time, the AWS DeepRacer League is the fastest way to get rolling with machine learning.
One of T-Systems’ very own – Jit Seah Sam Au, AI Sales Director at T-Systems Singapore – has joined the league, and swiftly advanced to the Regional Summit Qualifier 2 last October 19, 2022. A data science, artificial intelligence (AI), and cloud computing aficionado, Sam is an experienced director with a demonstrated history in regional sales, business development, and strategic partnerships.
Sam gained an early interest in AI back in 1994 when he studied at and graduated from the Nanyang Technological University (NTU), with a bachelor’s degree in Computer Engineering.
To expand his knowledge and build up his industry expertise, Sam then secured a Cyber Security and Data Science Specialist Certification from NTU Singapore, as well as an AWS Solution Architect Associate Certification in 2018. He started working on more complex AI, ML, and deep learning (DL) projects from 2017. Currently, he holds nine certificates from the NVIDIA Deep Learning Institute, mainly in the areas of machine learning and deep learning.
Sam is passionate about learning; he actively shares his vast AI knowledge and experience with curious learners. Currently, he mentors electrical engineering students at the National University of Singapore (NUS). In 2021, he served as a panel judge in the undergraduate NTU School of Computer Science and Engineering (SCSE) data science and artificial intelligence competition sponsored by AWS. He was also an adjunct lecturer in Temasek Polytechnic, teaching adult learners in the areas of Applied AI, Data Science for Business, AI Solutions Development, and Financial Analytics.
Joining the AWS DeepRacer League gave Sam the opportunity to learn more about AI, showcase his skill in training an ML model, and have the time of his life competing with fellow racers worldwide.
“It started with my great passion for machine learning, and I wanted to explore its more fun applications,” explains Sam. “Throughout the competition, there is an adrenaline rush as every race and the varied objectives in different tracks continue to test my personal discipline, adaptive learning, and knowledge.”
Personal fulfilment goals aside, Sam joined the race to expand his knowledge and show the deep AI expertise of T-Systems Singapore in action.
“It is self-fulfilling, definitely, and we get really excited as a community. But the competition not only consolidates my understanding of reinforcement learning, it also helps me in troubleshooting with detailed analysis,” explains Sam. “All these skills are transferable. It helps me a lot in my profession, especially when we are doing real machine learning for industrial applications.”
Sam started competing in May 2022 during the AWS ASEAN Summit and participated in the AWS DeepRacer Open Division. To better understand how he trained, let’s take a closer look at the virtual race car he is competing with—the AWS DeepRacer.
An AWS DeepRacer is a 1/18th-scale autonomous race car that serves as a platform for machine learning. Participants can learn the fundamentals in a free AWS Training and Certification course. Afterward, they get hands-on with reinforcement learning by participating in a workshop to build and train their models.
Reinforcement learning is a subset of machine learning and, as its name suggests, it is a training method built on rewarding desired behaviours. Through reinforcement learning, racers train their AWS DeepRacer to stay on track and go faster. Racers get tips, tricks, and tutorials on how they can train a model to achieve the fastest lap and manoeuvre increasingly challenging tracks.
But it’s not as easy as it sounds. Every model will be trained differently depending on the track, which increases in complexity over longer distances, sharper turns, and more objects to avoid. “To me, it is actually quite addicting, like gaming,” quips Sam. And as with mobile games, you also get new skins for your virtual race car for every track you conquer.
Racers like Sam train their model and reinforce desired behaviours, such as speeding up and turning a corner, through reward functions in code. Training a model can be quite similar to rendering a video: you wouldn’t really know the final outcome until you view the finished product hours later. And only then can you see where the training has gone awry, and you might need to redo your lines of code.
“It takes time to train a very good model,” explains Sam. “We analyse things, we see where our path is, we get insights from previous training, and then we have to benchmark—all of these things take time.” As such, small mistakes can be costly and time-consuming.
One good training session can take 10 to 24 hours, or even days, depending on the complexity of the model. “You can do trial and error, but one wrong move is going to cost you. So, you really have to be more clever about what you do and how you train,” cautions Sam.
In the AWS DeepRacer Virtual Circuit, racers start competing in the Open Division. Only the top 10% of racers get to advance to the Pro Division for a chance to compete in the Pro Finale. The top three racers in the monthly Pro Finales will qualify for the AWS DeepRacer League Championship Cup and earn a trip to Las Vegas to attend AWS re:Invent 2022, where racers will compete for the title and the grand prize. The event will be held from November 28 to December 2, 2022.
Winners at each Summit will qualify for regional playoffs to determine racers to advance to the AWS DeepRacer League Championship Cup. The Regional Summit Qualifier Races will take place in the AWS DeepRacer console, and broadcast live on the AWS Twitch channel. Participants will be separated into four groups by region, and one race for the public sector.
In 2022, the AWS DeepRacer League Summit Circuit returned to AWS Summit events worldwide so that racers can compete in person. Racers build autonomous remote control (RC) cars and train their models to conquer the new AWS Summit Speedway, the longest physical track in AWS DeepRacer history.
Sam participated in the AWS Summit Singapore 2022, his first physical race, where he got to meet the top racers and world champions in the league. In his first attempt, he made it to ninth place on the leaderboard. After getting feedback and valuable advice from top racers in the league, Sam seized fourth place in his second attempt. As part of the Winner’s Circle, he was rewarded with a physical AWS DeepRacer model and a limited-edition AWS DeepRacer League jacket.
Follow Sam’s results
Sam was among the top ten fastest racers in the ASEAN Qualifier Race of the AWS DeepRacer Summit Circuit, held in May 2022. Since then, he has moved his way up the competition:
While the championship event at re:Invent 2022 might be out of Sam’s reach, his passion and commitment to continuous learning powers higher performance. More than a test of speed, our participation in the AWS DeepRacer League is a showcase of our deep knowledge of AI and how it can be trained for intelligent solutions. Get in touch with our AI experts today to enable model training and automate scale-up into live environments.
