Multi-agent RAG, contract intelligence, and enterprise platform migrations. Sanitized architecture write-ups from real engagements at FPT Software and HTI Group.
LangGraph multi-agent system over hierarchical semantic chunks. AWS Lambda + ECS Fargate. Distributed observability via Langfuse. POC to enterprise production.
Migrated legacy HR conversational AI for a Fortune 500 chemicals corporation. Event-driven microservices on AWS, semantic caching, LLM-as-judge eval, zero-trust + PII redaction.
Skill-based framework with prompt chaining + Qdrant vector RAG. AWS Bedrock as multi-model LLM gateway. Async batch jobs on SQS scale to thousands of contracts per run.
On-premise RAG chatbot over the company document library. Hybrid Qdrant + MongoDB retrieval, hierarchical chunking, multi-LLM routing (GPT + Gemini), and self-hosted observability with Langfuse + Prometheus + Grafana.
GRI-aligned KG-RAG over Vietnamese bank ESG reports. Neo4j + hybrid retrieval + cross-encoder reranking + LLM-as-judge fact-checking. 88.14% accuracy on 1,440 expert-annotated QA pairs.
I bridge the gap between AI research and enterprise production. Over the past 3+ years, I’ve engineered scalable RAG pipelines, autonomous multi-agent systems, and zero-trust LLMOps infrastructures for multi-billion dollar corporations. My focus is moving AI out of the notebook and into highly available, compliant AWS microservices.
Previously built at & trusted by
Categorized by where it sits in the AI engineering stack.
Multi-agent systems, RAG patterns, model orchestration.
Production AWS infrastructure, IaC, container orchestration.
Observability, evaluation, compliance, performance.
Vector, full-text, graph search. Hybrid retrieval at scale.