Research
What I'm learning
Active areas: NLP, large language models, agentic AI, distributed systems, and MLOps. These are working notes, not finished essays — expect iteration.
All areasAgentic AIDistributed Systems
Agentic AIAugust 15, 2025
MCP and why it's the boring layer agents need
Working notes on the Model Context Protocol: what it standardises, where it shines, and the sharp edges I've hit building AutoDevs on it.
Read notesDistributed SystemsApril 3, 2025
Notes: checkpoints and fault tolerance in distributed training
A working summary of checkpoint strategies — full, async, and sharded — and why restart-mid-run still takes minutes.
Read notes