- 🎓 Education: Final-year B.Tech in Computer Science & Engineering (AI & ML) at Keshav Memorial Institute of Technology (KMIT).
- 🛡️ Final-Year Project: MCP-Shield — an inline semantic firewall for Model Context Protocol (MCP) traffic. Ships as a Python
@shielddecorator (mcp-shield-llm) with a 3-tier cascade (header check → vector/entropy filter → local fine-tuned SLM) to catch indirect prompt injection and confused-deputy attacks. - 🤖 Building: Agent-Dummie under Dummie-Labs — a personal, local-first AI companion with dual memory (SQLite + Obsidian vault), built on local LLMs via Ollama.
- 🧠 Exploring: Small Language Models for cybersecurity — training compact security-focused models (100M–1B params) for prompt-injection and phishing defense.
- 📚 Target: GATE 2027 (Data Science & Artificial Intelligence), then a Master's in AI.
- 🇩🇪 Sprachen: English, Telugu, and learning German (Ich lerne Deutsch!).
| Project | What it is |
|---|---|
| MCP-Shield | Security middleware for MCP: detects indirect prompt injection and confused-deputy attacks using a tiered cascade, down to a local SLM. |
| Agent-Dummie | A local AI companion with deterministic, gated memory, voice, vision and a memory-vs-search intent router. Work in progress. |
| PhishNinja | NLP-driven phishing and spam detection with explainable AI, shipped as a real-time security analysis mobile app. |
| AppCompiler + evals | LLM pipeline that compiles a plain-English request into UI, API and DB schemas, with an eval harness, tracing and CI. |
- IEEE (DELCON): Phishing Detection: A Comprehensive Survey of Methods and Trends
- Indian Patent Application (Published): PhishNinja: An NLP-Driven Framework for Phishing and Spam Detection Using Explainable AI — App. No. 202541111904 A (co-inventor)
- Next up: MCP-Shield research write-up
