/ AWS

AWS re:Invent 2019

Another late post, but that's just how it goes. I attended AWS re:Invent with some collegues back in December of 2019 in Las Vegas. The conference spans multiple hotels on the main strip and has something like 70-80k attendees. Since it was my first time attending, I wasn't really sure what to expect. Turns out, I had a lot of fun! There were lots of opportunities to try out different AWS technologies.

The first workshop I was able to attend was about using cloud technologies to deploy fleets of robots. I was able to use AWS RoboMaker to deploy an ROS application to a robot and control it remotely. There were some hiccups in the demo (Wi-Fi was terrible that first day), but I was actually the first person to successfully connect to my robot and get it moving! After the initial setup, transmitting program updates to multiple robots and activiting them becomes a trivial task. The robots can also benefit from using the cloud to handle any heavy processing. RoboMaker would be an interesting technology to investigate if you need to easily deploy applications to a fleet of robots and/or could benefit from the diverse capabilities of AWS services.

Next up, I attended a workshop on using Amazon SageMaker to train a blackjack AI with a neural network. It was less hands-on than I had been hoping, but was still really interesting. They set up cameras to take pictures of actual cards being played on a blackjack table. After tweaking their camera setup, they used the resulting pictures to train an AI to play blackjack. Some interesting learnings came about from this endeavor. Noise had to be introduced back into the images because they were too perfect. They used reinforcement learning to maximize the cumulative future reward. I was introduced to the concept of a Markov state, which occurs when a future state can be determined from a present state without past context. All good stuff. This project was really interesting to me and inspired me to make a text-based blackjack game, then train a model to make predictions about the outcome of a hand (coming in a future post).

There were several announcements that came out during the course of the conference. Some were pretty interesting, like Amazon Braket (apparently bra-ket is a notation for quantum states) for experimenting with quantum computing. I do not yet fully understand what can be done with quantum computing, but I would love to check it out and build something. It's currently in a preview state though, so you need to be selected to participate. Amazon Augmented AI provides human resources on demand to review machine learning predictions. Really useful for increasing confidence in your training processes. Then there is Amazon CodeGuru for automated code reviews and performance recommendations. It uses machine learning to determine issues in your code. Unfortunately for me, it only supports Java at this point in time, but it promises more languages coming soon.

They had two booth areas, one mostly for vendor advertising and one more for hands-on displays. The latter area was a lot of fun. I toured a smart home that had a mirror to help with shaving, an automated lawnmower, and even a cloud connected Karma Revero.

karma_revero

There was an AWS DeepRacer track set up for people to race autonomous cars powered by their machine learning models. I got there just in time to see a team fail gloriously as their car deliberately turned and shot off the track. There were also a bunch of tables set up for AWS employees to showcase their personal projects. My favorites were a drone that followed a remote control car on the ground, an automated garden integrated with Alexa, an autonomous boat that could navigate between two pillars, a robot opponent in an air hockey game, and a dinosaur safari using RoboMaker.

dinosaur_safari

The final event of the conference was the AWS re:Play party at the Las Vegas Festival Grounds. There was tons of free food and they had huge tents set up for various types of musical performances from rock to country to EDM. Outside, they had a dodgeball rink with real flames that shot out of the corners and live drummers. Another tent had all sorts of different activities, including an old school arcade, broomball, a light installation art thing, and the biggest bouncy house ball pit I've ever seen. To top it all off, AWS partnered with Intel to bring their Uplift drone light show to the skies above the fairground. Combined with some music from Kacey Musgraves, the drones made for a truly transcendent experience. It was a fantastic end to a fantastic week. I feel like AWS re:Invent is an experience that every software engineer should be exposed to at least once.

drone_show

Stan Nelson

Stan Nelson

I love programming. I'll create stuff for fun outside of work just because I like figuring out problems and the act of creating. I also enjoy cars, games, hiking, and going on adventures.

Read More