For anyone interested in getting their toes wet with deploying ML on small embedded devices, I cannot overstate how amazing Edge Impulse is (https://edgeimpulse.com).
Really interested hearing your experience with them! I worked on a project a decade ago on predictive analytics and it cost almost seven digits and required a decent amount of infrastructure overhead. Curious how this perform since everything runs on the board.
If nothing else, follow their guides for the arduino nano 33 ble sense + TinyML. They are a lot of fun!
The TinyML book by Pete Warden and Daniel Situnayake is an invaluable resource to getting started! Above all else I would recommend coming up with the top level details of a potential project as you work through the book' tutorials. Learning by doing (especially something one enjoys) is above all else the best way to truly grok something.
Most if not all of the weekly webinars and all the recordings from TinyML Asia 2020 and TinyML Summit 2021 are on this youtube channel: https://youtube.com/c/tinyML