Building AI agents: experiments, papers, and what actually works

Chennai · X @SAM_JOSHUA_S

I train and evaluate small coding and security agents (tool use, RLVR/GRPO, agent security). Honest numbers, not demos that only work in the screenshot.

Start here

• Sec RLVR Hub: the secure-code model family
• CVE Explainer: plain-English CVE triage

Interests: AI agents · LLM tool use · agent security · evals · multi-agent systems