Racing into the longer term: How AWS DeepRacer fueled my AI and ML journey

In 2018, I sat within the viewers at AWS re:Invent as Andy Jassy introduced AWS DeepRacer—a totally autonomous 1/18th scale race automobile pushed by reinforcement studying. On the time, I knew little about AI or machine studying (ML). As an engineer transitioning from legacy networks to cloud applied sciences, I had by no means thought-about myself a developer. However AWS DeepRacer immediately captured my curiosity with its promise that even inexperienced builders may become involved in AI and ML.

The AWS DeepRacer League was additionally introduced, that includes bodily races at AWS Summits worldwide in 2019 and a digital league in a simulated setting. Winners would qualify to compete for the grand champion title in Las Vegas the next yr. For 2018, as a result of AWS DeepRacer had simply been unveiled, re:Invent attendees may compete in particular person on the MGM Grand utilizing pre-trained fashions.

My colleagues and I from JigsawXYZ instantly headed to the MGM Grand after the keynote. Regardless of lengthy queues, we persevered, observing others racing whereas we waited. Members answered questions on driving preferences to pick a pre-trained mannequin. Not like later competitions, racers needed to bodily comply with the automobile and place it again on observe when it veered off.

We seen that the AWS-provided fashions had been unstable and sluggish by right now’s requirements, regularly going off-track. We concluded that rapidly changing the automobile on the observe may lead to an excellent lap time. Utilizing this technique, we secured second place on the leaderboard.

The night time earlier than the finals, we realized that we had certified due to a dropout. Panic set in as we realized we’d be competing on stage in entrance of hundreds of individuals whereas realizing little about ML. We frantically tried to coach a mannequin in a single day to keep away from embarrassment.

The subsequent morning, we discovered ourselves within the entrance row of the primary auditorium, subsequent to Andy Jassy. Our boss, Rick Fish, represented our crew. After an lively introduction from Indycar commentator Ryan Myrehn, Rick set a lap time of 51.50 seconds, securing the 2018 AWS DeepRacer grand champion title!

Picture 2 – Rick Fish accepting the AWS DeepRacer trophy from Matt Wooden

2019: Constructing a group and diving deeper

Again in London, curiosity in AWS DeepRacer exploded. We spoke at a number of occasions, together with internet hosting our personal A night with DeepRacer gathering. Because the 2019 season approached, I wanted to earn my very own finals spot. I started coaching fashions within the AWS DeepRacer console and experimenting with the bodily automobile, together with distant management and first-person view initiatives.

On the 2019 London AWS Summit, I received the AWS DeepRacer Championship with a lap time of 8.9 seconds, a big enchancment from the earlier yr. This occasion additionally sparked the creation of the AWS DeepRacer Community, which has since grown to over 45,000 members.

My curiosity in understanding the interior workings of AWS DeepRacer grew. I contributed to open supply initiatives that allowed working the coaching stack regionally, diving deep into AWS companies corresponding to Amazon SageMaker and AWS RoboMaker. These efforts led to my nomination as an AWS Neighborhood Builder.

Engaged on group initiatives improved my abilities in Python, Jupyter, numpy, pandas, and ROS. These experiences proved invaluable once I joined Unitary, an AI startup centered on decreasing dangerous on-line content material. Inside a yr, we constructed a world-class inference platform processing over 2 billion video frames each day utilizing dynamically scaled Amazon Elastic Kubernetes Service (Amazon EKS) clusters.

Picture 3 – Unitary on the AWS London Summit showcasing dynamically scaled inference utilizing 1000+ EKS nodes

2020-2023: Digital racing and continued development

The COVID-19 pandemic shifted AWS DeepRacer competitions on-line for 2020 and 2021. Regardless of this, thrilling occasions just like the AWS DeepRacer F1 Professional-Am stored the group engaged. The introduction of the AWS DeepRacer Evo, with stereo cameras and a lidar detector, marked a big {hardware} improve.

In-person racing returned in 2022, and I set a brand new world report on the London Summit. Whereas I didn’t win the finals that yr, the expertise of competing and connecting with fellow racers remained invaluable.

Photos 4 & 5 – the writer hoists the trophy from the 2022 London Summit (left) DeepRacer Neighborhood members and Pit Crew internet hosting a AWS DeepRacer workshop at re:Invent 2023 (proper)

2023 introduced extra intense competitors. Though I set one other world report in London, it wasn’t sufficient for first place. I ultimately secured a finals spot by successful a digital league spherical for Europe. Whereas my efficiency within the finals didn’t enhance on earlier outcomes, the chance to reconnect with the AWS DeepRacer group was rewarding.

Conclusion: The lasting impression of AWS DeepRacer

Over the previous six years, AWS DeepRacer has profoundly impacted my skilled and private life. It has helped me develop a powerful basis in AI and ML, enhance my coding abilities, and construct a community of buddies {and professional} contacts within the tech business. The expertise gained by AWS DeepRacer immediately contributed to my success at Unitary, the place we’ve achieved recognition as a high UK startup.

Because the official AWS DeepRacer league involves an finish, I’m excited to see what the group will obtain subsequent. This journey has formed my profession and life in methods I by no means anticipated once I first noticed that small autonomous automobile on stage in 2018.

For these concerned with beginning their very own AI and ML journey, I encourage you to discover the AWS DeepRacer sources obtainable on the AWS web site. It’s also possible to be a part of the thriving group on Discord to attach with different lovers and study from their experiences.

Concerning the writer

Matt Camp is an AI and ML fanatic who has been concerned with AWS DeepRacer since its inception. He’s at present working at Unitary, making use of his abilities to develop cutting-edge content material moderation know-how. Matt is an AWS Neighborhood Builder and continues to contribute to open supply initiatives within the AWS DeepRacer group.