What I'm doing now.

1. १

   Building

   AgentDiff, now in pre-release. Agents fail quietly. A prompt tweak or model update makes the agent loop, waste tokens or cost more, and the test still passes because the final answer looks fine.

   The storyagentdiff.app

2. २

   Working on

   First500days · AI Engineer (freelance). Building the AI API layer for AYUVA, a UK healthcare navigation platform, covering symptom triage, medical report interpretation, care journey generation, and a guarded health assistant..

   JRat’s Studio · Software Engineer (freelance). Developing secure, SEBI-compliant financial technology systems, optimizing high-performance data reconciliation pipelines, and architecting zero-trust AWS cloud infrastructure..

3. ३

   Writing

   Latest: Introducing AgentDiff: Catch Silent Agent Regressions Before They Ship

   Read it

4. ४

   Recording

   Out now: How Attention Actually Works: Math, GPU Memory, and Systems

   Next up: FlashAttention, KV caching, Systems-level LLM optimisation.

   Watch

5. ५

   Reading

   Attention Is All You Need · Paper

   Amazon Aurora: Design Considerations for High Throughput Cloud-Native Relational Databases · Paper

   Full readlist

6. ६

   Off the clock

   Painting and cooking, on weekends.

Inspired by the /now page movement. Want to work together on what's next? Say hello.