PSIML 11 recap

In August, Serbia hosted PSIML 11 — a summer school on machine learning for students from the country’s top universities.

My personal highlights this year:

1. The people are incredible. Extremely smart students, many of whom I would want to work with full-time right now.

2. We had five Russian-speaking lecturers covering topics from audio to agents. It’s great to see that my efforts to connect the Russian and Serbian ML communities are paying off.

3. I made merch for the first time — t-shirts, tote bags, thermoses, and even DIY mascot construction kits printed on 3D printers.

4. Yandex became a sponsor of the school this year and is even bringing us a rover to demo on sponsor day.

5. I mentored a project on diffusion models in audio for watermarking.

We even published the code on GitHub: github.com/martysai/psiml11-audio-with-diffusion

And the models on HuggingFace: AudioSeal robustness to diffusion attacks

I was also glad to prepare and give a lecture on Automatic Speech Recognition.

PSIML is an important non-profit educational initiative for Serbia, and I’m very grateful that I get to do a lot of good things for it.