At Deloitte, I build production GenAI platforms that turn natural-language requests into reliable, low-latency actions. My work spans LLM integration, agent orchestration, RAG, and event-driven backend services on AWS.
For Disney, I architected a real-time GenAI 3D control platform enabling live Unreal Engine scene control through natural-language prompts. The platform achieved sub-200 ms end-to-end latency, supported 50+ concurrent WebSocket sessions, and reduced scene-manipulation response time by 65%.
I designed a four-agent LangGraph orchestration layer with tool calling and 90%+ intent-classification accuracy across 30+ scene operations. I also integrated Amazon Bedrock Claude through LangChain to accelerate LLM integration and support multi-turn context-aware reasoning.
On Deloitte's AI Assist platform, I built core RAG, configuration, and LLM-service modules for 500+ enterprise users, rewrote vector search in Rust to more than double ingestion throughput, and shipped agentic workflows that reduced task completion time by about 35%.
