MS Computer Science @ UC Santa Cruz, building at the intersection of AI agents, enterprise infrastructure, and developer tooling. Previously interned twice at HPE/Juniper Networks, where I built AI agent systems and CI/CD automation for production engineering teams.

AI agents have unrestricted access to tools, file systems, and APIs with zero governance, which is why enterprises won't touch them. OpenClaw Gateway is an open-source MCP (Model Context Protocol) security proxy that sits between your LLM and its tools, enforcing enterprise-grade controls without changing a line of agent code.

• Audit logging — every agent action captured to PostgreSQL with full context, parameters, and results
• Multi-layer guardrails — regex pattern matching for known-dangerous commands (rm -rf, fork bombs, disk wipes) + LLM-based semantic analysis for novel threats
• Human-in-the-loop approval — high-risk actions paused and routed to a human reviewer via Slack, email, or webhook before execution
• Built with: Python FastAPI PostgreSQL LangGraph Docker

A privacy-focused alternative to cloud AI chat interfaces. Bring your own API keys, keep your data local, and get features that hosted tools don't offer.

• Multi-provider support: OpenAI, Anthropic, Google, Ollama — switch mid-conversation
• Conversation branching with a subchat system for non-linear exploration
• Multi-agent discussion mode: query models in parallel, orchestrate cross-model debates
• RAG-powered semantic search across your full conversation history
• PII redaction pipeline using local models to sanitize prompts before any cloud API call
• Built with: Python SQLite React FastAPI