Chien-Ting Hung

About

Hands-on LLM/RAG engineer with production experience improving retrieval quality and stability by addressing messy data, vector DB inaccuracy, and top-K issues—ultimately redesigning to hybrid search with tuned keyword/semantic weighting and MCP-based data supplementation. Also brings strong AKS/Kubernetes deployment experience, optimizing CI/CD speed via lightweight local Docker validation and decomposing pods to avoid full rebuilds, plus a metrics-driven approach to agent/workflow testing and traceability.

Experience

Education

Key Strengths

• Deep practical RAG troubleshooting (data quality, top-K behavior, retrieval accuracy)
• Improved retrieval accuracy by redesigning to hybrid search and tuning keyword/semantic weighting
• Performance optimization under load (index adjustments; dynamic top-K to manage query latency)
• Production-minded MLOps/infra experience with Docker, Kubernetes, and AKS
• Reduced deployment time by creating lightweight local Docker test environment and splitting pods to avoid full rebuilds
• Systematic agent reliability approach (explicit tool-call decisions, traceability, E2E tests incl. failure recovery, metrics + fallback tracking)
• Effective cross-functional collaboration with business stakeholders using value framing, visualization, and user-involved testing feedback loops

Reference Highlights

• Led AI team responsible for end-to-end architecture and development
• Thorough design process informed by market trends/frameworks
• Strong cross-functional communication with PM team
• Drives consensus in communications
• Able to make complex concepts easy to understand using visuals and examples
• Ensures releases occur only after testing verification
• Responsible
• Knows what he is doing
• Proactive in learning (reads AI papers to solve issues)
• Open to discussion
• Patient with non-technical stakeholders
• Explains technical issues clearly
• Uses examples/demonstrations to aid understanding
• Able to overcome technical hurdles (improved OCR accuracy through tuning)