CS student who'd rather ship an agent than talk about it. Right now that means MCP servers, LangGraph pipelines, and figuring out how to make LLM apps that don't fall apart the second someone red-teams them.
- SV-DLPS at DRDO, endpoint DLP system, 30+ file types, 96.8% detection across 300+ docs
- AgentShield, agentic security platform, LangGraph with 3 MCP servers and live WebSocket streaming
- RedTeamBench, automated LLM red-teaming with PyRIT and Garak, mapped straight to MITRE ATLAS
- AI doc-automation tool at Mishka Tech, Azure OpenAI with structured prompts and a human review loop
- 3 published papers (IEEE x2, Scopus x1)


