{{ p.name }}
{{ p.detail }}
AI researcher · AI engineer
Machines that reason,
machines that listen.
I'm Zeynel. I work on LLM post-training, agents, and speech processing at Codeway. My research interests include LLM reasoning and reinforcement learning, and I write about these topics at posttraining.dev.
01
Research
Looks right. Prove it.
02
What I do at Codeway
P(works | demo) ≠ P(works | users)
Speech for language learning
I build the voice experience of Learna: speech recognition designed for people still learning the language they're speaking.
LLMs & agents
Agents and LLM features for education products. The interesting problems live where research meets a real user.
Post-training & evaluation
Fine-tuning models after pretraining, then checking whether they actually got better. One good sample is an anecdote; an eval set is evidence.
03 · writing
posttraining.dev
My blog on what happens to a model after pretraining ends: reasoning, RLVR, distillation, reward models. Intuitions checked against real runs.
Read the blog ↗04
Talks & videos
From random walks to reasoning machines, out loud.
{{ talk.title }}
05
From taking integrals
to taking gradients.
Education & work
- mathematics
- probability
- learning
- RL policies
News
{{ n.date }}
{{ n.text }}
06
Beyond the terminal
Away from the terminal: chess games and conversations about research.