I design and build production-ready AI systems — from intelligent API gateways and autonomous agents to real-time data pipelines.
I focus on the layer between raw LLM capabilities and real-world applications: routing, memory, orchestration, and observability — the infrastructure that makes AI systems reliable at scale.
An enterprise-grade intelligent API gateway that semantically routes natural language queries to the right AI backend.
User Query → Semantic Router → casual_chat → LiteLLM (gpt-4o-mini)
→ code_assistant → LiteLLM (gpt-4o)
→ financial_quant → Celery + LangGraph Agent
Key capabilities:
- 🧠 Local embedding model routing — BAAI/bge-small-en-v1.5, sub-millisecond classification, zero API cost
- 🔐 Enterprise auth — SHA-256 hashed keys, Redis fast path + Postgres fallback,
expires_atenforcement - ⚡ Token bucket rate limiting — atomic Redis Lua script, per-user buckets
- 🤖 Autonomous LangGraph agent — live options data, risk-gated recalculation loop, 3-strategy selection
- 📊 Real-time dashboard — intent distribution, RPM sparkline, routing decision + confidence score
- 🔒 Production-ready — Nginx TLS, structured JSON logs, X-Request-ID tracing, Docker Compose full stack
Stack: FastAPI · LangGraph · Celery · Redis · PostgreSQL + pgvector · LiteLLM · Sentence-Transformers · Nginx
| Domain | What I Build |
|---|---|
| AI Infrastructure | Semantic routers, agent orchestration, LLM gateways |
| Backend Systems | Async APIs, task queues, real-time data pipelines |
| Security | API key management, rate limiting, auth middleware |
| Observability | Structured logging, distributed tracing, metrics dashboards |
Always building. Always shipping.