2022 started off no different from other years (well, at least compared with years in the COVID-era). However, things changed in early March when I was accepted into the AWS Community Builders program.
AWS Community Builder
The AWS Community Builders program offers technical resources, education, and networking opportunities to AWS technical enthusiasts and emerging thought leaders who are passionate about sharing knowledge and connecting with the technical community.
I was interested in the AWS Community Builder program as it related to my role at Arm, where I lead many of our technical engagements with Amazon (including AWS). Arm and AWS have a deep relationship around the Graviton EC2 family, which is comprised on Arm compute cores. However, I am more involved in IoT and Machine Learning, so I ended up in the Machine Learning community.
I planned to use this opportunity to share my experience in IoT and Machine Learning while expanding my knowledge of AWS services. I set up a new blog on dev.to and wrote my first post describing my goals. More on my involvement with the AWS Community Builders soon…
Multi-species BearID
In April, our second paper from the BearID Project, Multispecies facial detection for individual identification of wildlife: a case study across ursids, was published in Mammalian Biology. For this paper we collaborated with Russ Van Horn of San Diego Zoo Wildlife Alliance to train a multispecies bear face detector using images of bears living under human care. We were then able to use this multispecies detector to build a full pipeline for identifying Andean Bears.
AWS Summit San Francisco
Also in April, I attended the AWS Summit in San Francisco. Although I have been to many AWS events, this was my first as an AWS Community Builder. The AWS CB team hosted a great reception one of the evenings where I had a chance to meet my fellow community builders and the AWS team.
BearID at the Edge
In May, I completed an experimental edge deployment of BearID models trained in Azure Custom Vision using Azure Percept. The blog post on the Microsoft IoT Blog, Wildlife Monitoring and Conservation with Azure Percept, details the training and deployment.
Amazon re:MARS
In June I attended the Amazon re:MARS conference in Las Vegas. This was my second time attending this event which focuses on Machine Learning, Automation, Robotics and Space. I attended a lot of great sessions, including a workshop combining robotics and augmented reality (see the photo of the Jetbot and Magic Leap above). I managed to connect with other AWS Community Builders and a couple AWS Heroes, including Peter Hanssens:
Bearcam Companion
From April through September, I was working on an AWS-based web application, the Bearcam Companion. I documented the development from start to finish in the following 8-part blog series:
- Part 1: provide the background and define the M❤️P
- Part 2: use Amplify Studio to define the data model
- Part 3: develop the frontend with Amplify and React
- Part 4: add authentication and improve the UX
- Part 5: version control with GitHub and ML with Rekognition
- Part 6: storage with S3
- Part 7: Lamdas and Lanbda Layers
- Part 8: Deployment
You can see a demo of the application here:
In August, the Bearcam Companion blogs got some ❤️ from the AWS Developers account on Twitter:
IoT Builders Live
In September, I represented Arm as the guest on the first ever AWS IoT Builders live stream. I had a great time talking about IoT development with hosts Dan Gross and Nenad Ilic:
Imagine
The Edge Impulse Imagine conference was also in September. I attended the first day, which was live at the Computer History Museum in Mountain View, CA. The event featured presentation and demos of the latest innovations in edge machine learning. During the keynote, Edge Impulse announced “the ultimate AI for nature camera” which they are developing with Conservation X Labs and the Arribada Initiative. There was also a great Impact Panel on conservation and AI ethics, moderated by WILDLAB.NET’s Stephanie O’Donnell (there was even a BearID cameo in her intro slides):
October Accolades
October brought more accolades for BearID, starting with an article in The Daily Beast about the project and Bearcam Companion application. The article corresponded with Fat Bear Week. I was then interviewed on the AWS Twitch channel during a DeepRacer live stream [Edit: the stream is no longer available]. The Bearcam Companion blogs were promoted by the AWS Machine Learning account on LinkedIn. Finally, I was recognized as a runner up in the AWS Amplify x Hashnode Hackathon for the Bearcam Companion as detailed in Amplify-ing Bears.
Arm DevSummit
Also in October, I published a blog titled Accelerate IoT Development with Arm Virtual Hardware on AWS, leading up to the Arm DevSummit. Note: you can still watch content on-demand!
AWS Machine Learning Hero
In one of the top highlights of the year, I was selected as an AWS Machine Learning Hero in the final cohort of 2022. This came as a surprise, as I had only been selected to the AWS Community Builders in February.
I believe my work on the Bearcam Companion, the blogs and interviews, and most importantly, the amazing support from the AWS Machine Learning channel team, is what put me on the fast track to becoming an AWS Machine Learning Hero. But what is an AWS Machine Learning Hero? According to the AWS Heroes webpage:
AWS Machine Learning Heroes are developers and academics who are proficient with deep learning frameworks and are passionate enthusiasts of emerging Amazon ML technologies. They enjoy helping developers of all machine learning proficiencies learn and apply ML, at speed and scale, through Hero blog posts, videos, and technical sessions, as well as direct engagement.
You can read about myself and my 6 fellow AWS Hero inductees on this blog post. You can also find my card (shown above) here.
AWS re:Invent
The last major activities for 2022 took place in Las Vegas at AWS re:Invent. This was my 4th re:Invent, but the first as an AWS Hero. I must say, being a Hero at re:Invent comes with some serious perks. From the Heroes dinner to special lounge access and front-row seats for a shout-out from Werner Vogels during his keynote, it was quite the experience.
I had also been selected as a PeerTalk Expert. PeerTalk is a new onsite networking program for AWS re:Invent attendees. From the AWS Events app, you can connect with other attendees and set up face to face meetings. As an expert, I was available for connections and meetings and helped with promotion. I had a few great meetings through this tool.
My main function at re:Invent was representing my employer, Arm. Our team manned a small booth providing information about Graviton and our IoT solutions. I helped set up a Software Defined Camera demo which showed the same containerized software stack consisting of a vision based ML model communicating to AWS IoT Core through Greengrass running on 2 different physical boards and Arm Virtual Hardware.
As the culmination of my Bearcam Companion work for this year, I presented a DevChat in the Developer Lounge in the re:Invent Expo. The presentation covered background on the BearID Project, the Explore.org bearcams and the application I built using various AWS services. Sadly the presentation wasn’t recorded (perhaps I will record something similar and post it at a later date). The following day I was interviewed on the AWS Build On Live stream by the amazing Linda Haviv. This stream is still available (for now):
Migrating (Back) to Medium
I plan to utilize [Medium](https://bluevalhalla.medium.com/) as my primary blog moving forward. I have been posting and crossposting on Medium for years, so you can find many of my posts and follow me at [bluevalhalla.medium.com](https://bluevalhalla.medium.com/). I will cross post on my [dev.to blog](https://dev.to/bluevalhalla) when it is relevant.
Looking Ahead to 2023
Overall, 2022 was a big year for me, but I’m already looking forward to next year. The BearID Project team will be hard at work extending our application to work with trail camera video clips (look for a new paper in the coming year). I will continue to extend the Bearcam Companion by using the 2022 data to build models to automatically identify the bears on camera in 2023 (with corrections from the bearcam community!). At Arm I will continue to focus on IoT and Machine Learning at the edge and hope to publish more content utilizing AWS services (who knows, maybe I’ll even tie it in with BearID).
Wishing you all a happy and prosperous 2023!
Follow me at bluevalhalla.medium.com!
