I research and train LLMs for what the industry needs
The full LLM post-training cycle: generating thousands of RL environments and training data, SFT, RL, evaluation, and a paper at an A* conference.
I train language models and build products on them

years of experience
products built
people on the largest team I led
I work both as an engineer on a team and as a turnkey contractor.
The full LLM post-training cycle: generating thousands of RL environments and training data, SFT, RL, evaluation, and a paper at an A* conference.
Agents, RAG, integrations — so that it holds the load and brings business impact.
From the first conversation to a finished service. A team of 2–7 people, 2–8 weeks.
I love technically hard products and problems in NLP and LLMs.
Turnkey in 2–8 weeks.
My expertise is backed by research papers.
We consolidated traffic from 200+ internal applications onto a single self-hosted 32B model: a separate GRPO expert per axis, merged with two-stage SLERP.
Products I built from scratch.
My expertise is backed by contests and competitions.
I share what I know in talks, and keep learning at workshops and schools.