Previous ideas:

  • Detecting gear faults (missing tooth, chipped tooth, cracked tooth, teeth wear) using tinyML to interpret vibration signatures from a drivetrain diagnostics simulator.
  • Baselining and improving aggregation algorithms in federated learning.
  • Running machine learning algorithms on camera traps to identify individual mammals within a larger population in remote locations.

Today:

  • An open-source machine learning pipeline to train use case specific end-to-end spoken language understanding (E2E SLU) models for resource constrained devices.

If you want to get a sense of why I've changed my research topic a few times, check out my previous post. There’s no need to dive into the specifics here (but I'm happy to chat about it over a beer!).

Do I know much about automatic speech recognition (ASR), speech-to-text (STT), natural language processing (NLP), natural language understanding (NLU), spoken language understanding (SLU)…? Nope! Do I know the exact direction of my research? Not yet! Do I know this direction is going to work out? Absolutely!

Rather than waste your time being stressed over making the right decision, make the decision right. – Dr. Ellen Langer (on The Rich Roll Podcast, episode 813)

I know it’s going to work out because it’s time to lean into uncertainty, adaptability, and flexibility. It’s time to stop f’ing about and make the decision right. Let’s go. 

This podcast episode is where I pulled the quote above from. If you're interested in listening to a fascinating conversation about how we might be able to think our way to chronic health and the connection between the mind and body, check it out.