Hi, I'm Aayush Kumar

Senior ML Engineer specializing in production AI security automation and RAG/agent systems. I build reliable LLM-powered workflows that ship under real constraints—low false positives, tight SLAs, and measurable operational impact.

Most recently, I delivered an agentic security automation pipeline that held false positives under 3%, safely auto-merged 30%+ of remediation PRs, and processed 50+ PRs/week within a 24-hour SLA using retrieval-grounded reasoning, calibrated decisioning, and guardrails.

Previously, I led an LLM-assisted incident intelligence system that reduced hallucinations by 18%, improved detection by 30%, and cut infrastructure overhead (220GB → 52GB memory, 6× lower compute costs) through systematic profiling, optimization, and evaluation-driven iteration.