AI Systems Engineer | Agentic AI & Multi-Agent Architect
Exploring the boundary between autonomous machine intelligence and core software architectures. Currently engineering production-grade Retrieval-Augmented Generation (RAG) ecosystems, real-time voice intelligence, and scalable AI infrastructure.
- 🛠️ Juggling multi-agent race conditions, optimizing vector storage retrieval, and debugging runtime environments at 2 AM.
- 🧪 Core Focus: Advanced Agentic AI frameworks (LangGraph, CrewAI), Model Context Protocol (MCP), and real-time audio streaming infrastructure.
- 🔬 Research & Collaboration: Associate at Next Tech Lab, collaborating on deep learning pipelines and autonomous intelligence architectures.
Dockerized Voice AI Platform engineered for ultra-low latency, production-ready real-time interaction.
- Implemented a cutting-edge LightRAG architecture for lightning-fast graph-based content retrieval.
- Leveraged LiveKit audio streaming and FastAPI to build high-performance audio inference pipelines.
- Fully containerized with Docker and deployed across AWS EC2/S3 infrastructure.
Context-Aware Voice Intelligence System built to optimize institutional query handling.
- Designed a dual-RAG routing pipeline that dynamically switches between unstructured PDFs and structured JSON sources depending on intent classification.
- Implemented complex session persistence mechanics enabling full conversation memory and graceful reconnection handling.
- Stack: Python, LangChain, Qdrant Vector DB, LiveKit STT/TTS.
Domain-Bounded RAG Engine built for AR-based cultural heritage reconstruction.
- Built a strictly bounded, domain-specific RAG pipeline to power immersive, factual contextual lookups in 3D historical environments.
- Handled context-injection barriers to avoid LLM hallucinations regarding historical facts.
- Multi-Agent Orchestration & Graph-Based Executions (LangGraph, CrewAI)
- Advanced Retrieval Engineering (GraphRAG, Hybrid Search, Dense/Sparse Routing)
- Real-time Streaming Infrastructures & Voice-to-Voice AI Engines
- Hardware-Software Co-Design & High-Performance Data Structures
