AI Code Reviewer (Multi-Agent System)
- Designed & built a production-ready, scalable system on Horizontally Scalable AWS EC2 using FastAPI and Producer-Consumer architecture
- Orchestrated LangGraph Multi-Agent workflows for sequential file indexing (PGvector) and deep code analysis
- Developed specialized agents with tools for Multi-Repo referencing, symbol & file retrieval, RAG-based context injection and agent task delegation
- Implemented Interactive Repo Reviewer: A GitHub chat agent for vulnerability discovery via hybrid search (SQL & Vector)
- Integrated Summarizer Agent to report fixed code vs. new vulnerabilities and suggest repairs directly on GitHub
- Designed a React.js UI for full-detail vulnerability management and visualization
- Solved Long-Context issues via intelligent tool-usage and context summarization
- Engineered secure cross-repo referencing with strict permission handling and data isolation
- VP of SE: "Output is quite good. It found some hidden subtle bugs"
- Reported 100+ risky vulnerabilities not caught by Claude Code; achieved 7x speed increase
- Optimized AI agent workflows, reducing token usage by 40% while increasing throughput
Blockchain Security Software Development
- Primarily blockchain transaction monitoring (NestJS, Microservice)
- Monitoring txns for 1.2M+ monthly active users (of 100M+ users)
- Impl. software to enable high-speed communication with blockchain nodes
- Impl. software to track company deposits/withdrawals and trading volume
- Maintain and deploy 100+ repos 100% automatically with Claude Code Agents
- Responsible for heavy AI automation
- Prometheus, Grafana
- Successfully reported double spending and keep trading volume in check
- New chains are completed in 1 day with Claude Code and custom Agents
Common Tools
- AWS, PostgreSQL, Kubernetes, Docker Compose
- System design
- Note: Development includes Software Design, Implementation, Testing, Deployment