Status Updates
Once a month I try to sit down and think about what has happened.
-
2026, March - Retreats, Meetups, and Beyond the Academic Game
-
2026, February - Agents, Fragmentation, and a Conference Talk
-
2026, January - MLOps Teaching Assistant and outreach
-
2025, December - EurIPS, Course development and holidays
-
2025, November - Early Project Update: Model Compression and Edge ASR Directions
About the project
It is titled “Edge deployment of deep neural networks via Model Compression for Healthcare Applications” and is a collaboration between Laerdal Medical and the Technical University of Denmark. It’s a 3-year PhD going from October 1st 2025 to September 31st 2028.
Here are a few key statements that motivates me about the project:
- We’d like to see how much performance we can squeeze onto devices with limited compute.
- Huge amounts of resources have already one into training models. Let’s make sure that they’re put to use by making them applicable on a wide range of devices.
- Centralizing AI inference is the standard but it’s not always clear winner. Distributing inference compute to the client-side can be beneficial, perhaps even necessary, due to privacy, latency or cost at scale.
Advice I keep returning to
- Andrej Karpathy: A Survival Guide to a PhD
- Tim Dettmers: Lessons Learned from Successful PhD Students
- John Schulman: An Opinionated Guide to ML Research
- Eugene Vinitsky
- Richard Hamming: You and Your Research
- Michael Nielsen: Principles of Effective Research
- Sam Altman: How to be successful
- Maxwell Forbes: The PhD metagame