I am an Undergraduate Computer Engineering Student focused on Low-Level Networking, Systems Security, and AI Agent Security Architectures. I specialize in building custom network proxy tools, asynchronous concurrency engines, and offensive/defensive pipelines specifically tailored for modern LLMs and Agentic AI applications.

• 🔭 Current Focus: Asynchronous traffic interception (asyncio/mitmproxy), Agentic Security (RAG & Self-Critique loops), and low-latency ML anomaly detection for HTTP/HTTPS packets.
• 🛡️ Featured Ecosystem:
  • llm-inspector — An enterprise-grade MITM proxy & interactive interceptor (Burp Suite-style) for auditing LLM/AI Agent endpoints with dynamic TLS termination and WebSocket control.
  • payload_generator_agent — An offensive Red-Teaming AI agent using RAG (OWASP Top 10 for LLMs / PortSwigger) and Reflection patterns to generate and self-critique attack payloads.
  • llm-attack-detector — A high-throughput ML pipeline (Logistic Regression + Hard Negatives) for real-time header & body payload classification.
• ⚡ Fun Fact: I strongly prefer official upstream package repositories over local mirrors for my Linux setups!

• LLM Inspector Proxy Core: Implemented TLS interception, asynchronous WebSocket event bus, and interactive packet interception UI.
• Agentic Red Teaming (Reflection Phase): Built self-critique (reflect_and_generate) loops using local LLMs (Ollama) to evaluate attack plausibility before human review.
• RAG Pipeline Integration: Indexing OWASP Top 10 for LLM Applications and PortSwigger research for context-aware payload synthesis.
• Closed-Loop Feedback Retraining: Feeding successful agent bypass payloads directly back into the llm-attack-detector-training dataset to continually harden the ML classifier.