Rudratosh Shastri

10+ years in backend. Python (FastAPI, Django), PHP (Laravel), SQL. Microservices, ETL and data pipelines, AWS, Docker. Lately LLMs and AI agent tooling.

How AI agents get tricked through the stuff they read. Prompt injection, guardrails and agent security, mostly by building benchmarks and breaking things on purpose.

taintgate, a small policy gate for AI agent tool calls, and buried-injections, a benchmark of 10 prompt-injection detectors on real agent attacks.

Chats about backend architecture, scaling data-heavy systems, or AI agent security. Happy to review guardrail ideas or collaborate on open-source. Just say hi!