Inside the model
Transformers, attention, fine-tuning, LoRA, RLVR, GRPO, reasoning, inference, and what newer models make possible.
From tokens to tools
I build production AI systems and write about how they work.
Clear, practical writing about model behavior, post-training, and the engineering behind dependable AI agents.
AI Tech Lead · builder · writer
Two tracks, one system
I write about models and the systems around them.
One track looks inside the model. The other looks at the engineering that turns model capabilities into reliable work.
Beyond the prompt
Sandboxes, memory, evals, tools, background agents, context engineering, authentication, and production reliability.
Now writing
What I’m working on.
I start with a practical question, show how the system works, and leave you with something you can build or test.
Agent systems · In progress
Agent Computers: Why Sandboxes Unlock Powerful AIFrom a chat window to a safe, stateful computer.
LLM systems · Up next
Attention, Without the Wall of MathA visual explanation of the mechanism behind transformers.
Model × agent systems
How GRPO Teaches Models to ReasonRewards, verifiable outcomes, and the DeepSeek connection.