ML Researcher & Builder
Bartosz Dzionek
I do machine learning research at Exa, where we’re building search that’s actually made for AI. Previously at Google. I care about the full stack — from the research that makes a model better to the engineering that gets it in front of people.
What I work on
Search & retrieval for AI
Building search that’s designed for models, not just people — the retrieval layer that lets LLMs reason over the live web.
Language models
Language modeling and the mechanics underneath it. My undergrad thesis dug into the low-rank softmax bottleneck.
Research → production
I like owning the whole arc: from the experiment that proves an idea to the system that ships it to real users.
Selected writing
All posts →- Hello, world (and a feature tour) A starter post that shows off the things this blog supports — LaTeX math, syntax-highlighted code, and embedded images.
- Search built for models, not people Web search was designed for humans skimming ten blue links. AI systems want something different — and that gap is the interesting problem.
- The low-rank softmax bottleneck, explained Why a language model’s final softmax layer can quietly cap how expressive its predictions are — and what you can do about it.
Selected projects
Low-rank softmax bottleneck
Undergraduate thesis on a known limitation of the softmax in language models, and ways to alleviate it.
LLMsLLM rap generation
A large language model fine-tuned to generate rap lyrics, with code reproducing the results.
Haskell · SearchQuoridor AI
A game-playing engine for Quoridor, written in Haskell. Search, heuristics, and a lot of pruning.
Applied MLML for conservation
Accessible deep learning for analysing wildlife camera-trap data — making field ecology a little less manual.