Summary
AI Engineer with 3+ years at FPT Software AI Center, delivering production LLM applications, agentic AI platforms, and MLOps infrastructure across enterprise and industrial domains. Applies AI across the full development lifecycle — from automated research and agentic planning through AI-assisted implementation, testing, and code review to real-time monitoring agents embedded in delivery workflows. Specialized a large-scale MoE LLM for the mainframe domain through a multi-stage alignment pipeline. Published researcher on LLM-based mainframe modernization; creator of MainframeBench, the first COBOL/JCL comprehension benchmark. Two-time international AI innovation award winner (IT World Awards Gold Globee; Asia-Pacific Stevie Gold Awards). Active contributor to Vietnam's AI ecosystem as Lab Coach at VinGroup's AI20K program.
AI Engineer and Researcher at FPT Software AI Center specializing in large language models, agentic systems, and applied ML for software engineering tasks. Published work on LLM-based mainframe modernization introduces MainframeBench — the first benchmark for COBOL/JCL/PL/I code understanding — with significant accuracy and summarization improvements covered by global AI media. Conducted a systematic alignment research campaign on a large-scale MoE LLM (SFT → SLERP → ORPO → DAPO); contributions include a novel MoE-aware SLERP merging technique. Research interests span LLM alignment methodology, multi-agent system design, and scalable AI infrastructure. Recipient of two international AI innovation awards (IT World Awards Gold Globee; Asia-Pacific Stevie Gold). Currently mentoring AI practitioners at VinGroup's AI20K program.
LLM and Agentic AI Engineer with 3+ years at FPT Software AI Center, delivering production AI systems end-to-end — from MoE LLM specialization and alignment to multi-agent orchestration, hybrid RAG pipelines, and agent harness engineering at enterprise scale. Applies AI across the full development lifecycle: automated research, agentic planning, AI-assisted implementation and testing, code review, and real-time monitoring agents embedded in CI/CD pipelines. International award-winning engineer with a published paper on LLM-based legacy code modernization. Currently coaching next-generation AI engineers at VinGroup's AI20K program.
AI researcher and engineer specializing in LLM alignment, empirical evaluation, and applied ML for code intelligence. Conducted a systematic alignment research campaign on a large-scale Mixture-of-Experts model — spanning preference optimization (ORPO) and reinforcement learning from verifiable code execution rewards (DAPO) — and developed a novel MoE-aware SLERP merging technique to resolve catastrophic forgetting. Published author on LLM-based mainframe code intelligence; creator of MainframeBench, the first benchmark for COBOL/JCL/PL/I code understanding. Research interests span LLM alignment methodology, RL from verifiable feedback, and multi-agent system reliability. Applies AI across the full engineering and research lifecycle — from automated literature synthesis and agentic planning to real-time monitoring — enabling self-improving, continuously validated workflows.
AI researcher and engineer with 3+ years building large-scale LLM systems, multi-agent frameworks, and AI infrastructure at FPT Software AI Center. Specialized a 20B-parameter Mixture-of-Experts model through a full alignment pipeline (SFT → SLERP → ORPO → DAPO), contributing a novel MoE-aware component-wise merging technique and applying RL from verifiable code execution rewards. Designed production-grade multi-agent orchestration with custom tool ecosystems, context management, and reliability guardrails — spanning foundation model fine-tuning, agentic platform engineering, and AI-driven development workflows. Published researcher on LLM-based code intelligence; creator of MainframeBench. Two-time international AI innovation award winner.
Work Experience
-
AI-Driven Engineering Practice: Transformed the engineering and research workflow by embedding AI agents across the full development lifecycle. Automated literature and technology research to accelerate decision-making; used agentic planning for task decomposition and sprint structuring; applied AI-assisted coding, test generation, and code review in daily practice. Deployed real-time monitoring agents that detect, diagnose, and respond to application issues as they occur, and integrated AI into CI/CD pipelines to enable continuously validated, self-improving delivery.
-
Legacy Modernization / XMainframe (2023–Present): Led development of an agentic AI platform for mainframe code analysis (COBOL/JCL/PL/I). Designed multi-agent orchestration with custom static-analysis tools and real-time streaming interfaces. Engineered a production-grade agent harness with context management, tool-result budgeting, middleware hooks, prompt caching, and reliability guardrails — informed by a systematic review of leading agent frameworks. Co-authored the published technical report.
-
XMainframe LLM Specialization (2025–2026): Specialized a large-scale Mixture-of-Experts LLM for mainframe domains through a 4-stage alignment pipeline (SFT → SLERP → ORPO → DAPO). Developed a novel MoE-aware SLERP merging technique to resolve catastrophic forgetting in expert layers, with meaningful gains in multilingual reasoning and code summarization. Built a domain-specific evaluation suite from real mainframe codebases. Also fine-tuned COBOL-Coder-14B on domain-specific data, achieving state-of-the-art performance on COBOL code generation and translation benchmarks.
-
AgentVista / Metis (2025): Architected an enterprise agentic AI platform with state-machine-based workflow orchestration and hybrid RAG pipelines combining vector search and knowledge graphs for code review, documentation, and knowledge extraction.
-
SoundAI & SenseAI – Time Series MLOps (2022–2023): Delivered a serverless MLOps platform for acoustic anomaly detection in manufacturing, including stream and batch pipelines, model operations, and automated deployment across Edge and Cloud environments.
-
TBA Xray i2 (2022–2023): Built industrial computer vision models, rule-based inspection logic, and edge-deployable AI applications for X-ray quality inspection in FPT's first project with LandingAI; delivered on-device inference pipeline running on NVIDIA Jetson hardware at production throughput.
-
Fine-tuned domain-adapted LLMs on curated datasets; optimized inference serving for production-scale throughput and latency requirements.
-
Built multi-stage data and model pipelines; established experiment tracking and model registry workflows adopted across AI Center projects.
-
Fine-tuned domain-adapted LLMs on curated datasets; optimized model serving infrastructure for production-scale throughput and latency.
-
AI20K is a competency-based AI training program at VinGroup designed to develop practical skills in building and deploying AI Agents in enterprise environments.
-
Coach and mentor practitioners across three competency tracks: AI Business & Product (problem framing, ROI, governance), AI Infrastructure & Data (data pipelines, MLOps/LLMOps, monitoring), and AI Application (LLM engineering, agentic design, RAG, multi-agent systems, evaluation).
-
Coach practitioners across three tracks — AI Business & Product, AI Infrastructure & Data, and AI Application — covering LLM engineering, agentic design, and multi-agent system deployment for VinGroup's enterprise AI program.
-
Facilitate lab sessions combining hands-on exercises, project builds, and mentor reviews across the Foundation, Specialization, and Enterprise Practicum phases.
Education
- Competitive research residency focusing on applied AI and production ML systems
- Trained and mentored by PhD-level researchers and global AI experts
Publications
Anh T. V. Dau, Hieu Trung Dao, Anh Tuan Nguyen, Hieu Trung Tran, Phong X. Nguyen, Nghi D. Q. Bui
- Developed a domain-specialized LLM for COBOL/JCL comprehension with significant gains in code understanding and summarization accuracy
- Created MainframeBench: first comprehensive benchmark for mainframe code understanding across COBOL, JCL, and PL/I
- Global media coverage: MarkTechPost (US), ITmedia AI+, Ledge.ai (Japan)
Honors & Awards
Best Rookie of the Year 2023
FPT Software AI Center
Hanoi, Vietnam
2023
Gold Globee Winner for Manufacturing
IT World Awards
International
2023
Gold Award for Innovation in Construction, Manufacturing or Materials
Asia-Pacific Stevie Awards
Asia-Pacific
2023
Competitions & Projects
Built an end-to-end sticker generation system with Stable Diffusion and optimized Triton deployment.
Developed a deep learning classifier for COVID-19 detection from cough audio and placed 3rd in Phase 1.
Built a credit scoring model using WoE-IV feature engineering and ensemble methods; ranked in the top 4%.
Certificates
Databricks Certified
Data Engineer Associate
2023
Udacity Nanodegree
Machine Learning DevOps Engineer
2022