Production microservices, multi-provider LLM orchestration, and on-chain apps.
AI sales-intelligence platform: multi-owner RBAC, idempotent assignment service, 8-phase zero-downtime migration, multi-provider LLM coaching.
AI career platform across a 5-service architecture. GPT-4o + Gemini fallback, conversational data extraction, real-time Socket.IO + SSE.
Multi-tenant AI education platform: institutional RBAC hierarchy, per-student token quotas, cost-analytics, scoped agent API keys.
Multilingual RAG assistant over 40+ languages on a LangChain pipeline with Gemini + OpenAI fallback and MongoDB backup infra.
Multi-agent trip planner: Gemini function-calling routes destination, itinerary & budget agents with code-enforced behavioral rules.
Personal link-management REST API in Go: tag search, automatic metadata extraction, JWT auth, multi-container Docker deploy.
The demo is the easy part. I obsess over the systems underneath — the access checks, the idempotent writes, the migrations that ship with zero downtime — because that's what makes an AI product survive contact with real users.