AI Engineer · AI Researcher
Computer Vision · Multimodal AI · AI Systems · Smart Manufacturing
About · Featured Work · Publications · Toolkit · Education & Honors
I build AI systems across research and engineering, from mathematical formulation and controlled experiments to backend runtimes, telemetry infrastructure, and reproducible pipelines. I care about systems that can be inspected, audited, and deployed reliably.
- 🔬 Research Assistant @ AiTA Lab, FPT University (Oct 2024 - Present).
- 🎓 Completed academic requirements for B.Eng. in Artificial Intelligence & Data Science at FPT University (degree conferral expected Nov 2026).
- 🏭 Current research focus: Adaptive Industrial Visual Intelligence (LG-IVI), specializing in language-grounded defect localization and inspection under zero- and few-shot constraints.
- 🧩 Areas of interest: Multimodal representation learning, VLM active learning, agentic research workflows, and IIoT/MES platforms.
🧪 LexiChem - Research and end-to-end AI system
Completed Capstone · First-author APWeb-WAIM 2026 Paper
- Natural-language to 2D/3D molecular generation through shared-latent cross-modal alignment in SELFIES space, evaluated on the L+M-24 benchmark.
- Data preprocessing pipeline across 693,090 training examples with RDKit canonicalization and SELFIES conversion.
- Multi-model serving pipeline with NVIDIA Triton Inference Server, RDKit chemical validation, and FastAPI backend.
- 🔗 Read Case Study
🏭 NexOps - Smart-manufacturing systems engineering
Working Prototype · AI & Simulation Lead
- Local-first MES/SCADA runtime connecting machine telemetry, production schedules, alarms, and shift OEE.
- Deterministic machine/shift simulations, MQTT pipelines via EMQX to FastAPI to TimescaleDB, and local Bambu Lab 3D printer hardware control over LAN.
- 🔗 Read Case Study & Demo
Pre-alpha · Personal System
- Human-in-the-loop research workspace designed to make scientific progress traceable and reproducible.
- Treats evidence, provenance, SQLite case state, checkpoints, and bounded agent review as first-class primitives.
I currently have five accepted or published papers spanning handwritten mathematical expression recognition, multimodal alignment, and active learning with vision-language models:
| Year | Paper | Venue | Role | Artifacts |
|---|---|---|---|---|
| 2026 | LexiChem: Shared-Latent Alignment for Faithful Text-to-Molecule Generation in SELFIES Space | APWeb-WAIM | First author | Case Study |
| 2026 | CorrTie: Correction-aware Tie-breaking for Active Learning with Vision-Language Models | APWeb-WAIM | Co-author | Paper · Code |
| 2026 | ChemAligner-T5: A Unified Text-to-Molecule Model via Representation Alignment | ICCIES | Co-author | Paper |
| 2026 | Enhancing Handwritten Mathematical Expression Recognition with Convolutional Block Attention Refinement Module and Multi-Task Learning | JIT (Taylor & Francis) | Co-author | Paper · Code |
| 2025 | Mask CoMER: Enhancing Handwritten Mathematical Expression Recognition with Masked Language Pretraining and Regularization | ICDAR (LNCS) | Co-author | Paper · Code |
View Google Scholar Citations →
| AI & Research | Python · PyTorch · PyTorch Lightning · Hugging Face Transformers · scikit-learn · OpenCV · RDKit · Weights & Biases · MLflow · NVIDIA Triton Inference Server |
| Backend & Data | FastAPI · Pydantic · Pandas · PostgreSQL / TimescaleDB · MongoDB · Redis · MQTT / EMQX · REST APIs |
| Engineering & Reproducibility | Docker · Docker Compose · Linux · Git · GitHub Actions · Nginx · Playwright · LaTeX |
- B.Eng. in Artificial Intelligence & Data Science, FPT University (Graduation expected Nov 2026)
- Capstone: LexiChem (Text-to-molecule generation workbench and multi-model inference serving)
- AI VIET NAM (AIO 2024): Graduate of the 4-tier AI Fellowship (Python, Machine Learning, Deep Learning, Generative AI & MLOps)
- AiTA Lab: Research Assistant (Oct 2024 - Present)
- 🥈 Silver Medal: Vietnam National Open Mathematics Olympiad for High School Students (2021)
- 🎓 50% Tuition Scholarship & Project Grants: FPT University, SIHUB Innovation Quest, RESFES HCMC
Research rigor, engineering ownership, and systems that survive beyond the notebook.
Get in touch: ngminhkhoayj2706@gmail.com · Explore more at purin1410.github.io