Hi, I’m Niels uit de Bos.
I’m currently looking for full time work in AI research engineering or AI/ML-related SWE work. Send me a message (niels@uitdebos.com) if you have any opportunities or would like to discuss potential collaborations! I have more than 5 years of professional software engineer experience (about 1 year of AI research engineering, and 1.5 additional years of applied ML), and I have a Ph.D. in mathematics (algebraic geometry).
This is a short overview of what I’ve been up to and what I’m interested in:
- AI research engineering: at the start of 2024, I did a research internship on AI alignment research (specifically mechanistic interpretability) as part of MATS 5.0 and 5.1, mentored by Adrià Garriga-Alonso. Some of the work I did there: Adversarial Circuit Evaluation (workshop paper at the ICML Mechanistic Interpretability workshop; tweet thread) and An evaluation of circuit evaluation metrics (LessWrong Post). I also collaborated on this paper that shows how Kolmogorov-Arnold networks with piecewise linear activation functions can be converted into ReLU networks and vice versa.
- Software engineering: I’ve been programming since I was 12, and have now done it professionally for about 5 years. I’ve seen and appreciated different extremes of programming style: very robust and well-crafted code at Zopa and G-Research (finance) on one end, and very quickly written and huge value providing code at Y-Combinator ML-startup Depict. And recently I’ve been running AI experiments, which requires yet another style of coding with different trade-offs.
- Mathematics: I did a Ph.D. in algebraic geometry (geometric Langlands) at the University of Duisburg-Essen advised by Jochen Heinloth.
- Formal verification: Recently I’ve been working on proving in Lean when you can merge two tensors views (shape/stride pairs) in tinygrad (it’s the same in PyTorch). During the pandemic I studied Coq (AKA The Rocq Prover) with a group of friends, and we worked on this.
- Linguistics: I don’t only like programming languages, I also like natural languages a lot! I know many things about many languages, won the Dutch computational linguistic Olympiad, speak Dutch, English, French, German, know a fair bit of Latin, Ancient Greek, Swedish, Hindi, Esperanto, and am currently learning Mandarin Chinese.