About me - Thien Do
I’m passionate about optimizing AI systems and making them run faster while maximizing GPU utilization. I’m always curious about how to squeeze every bit of performance out of hardware and eliminate waste in AI infrastructure.
My expertise spans the full AI engineering stack - from research and modeling to production deployment and optimization. I actively contribute to open source projects like Hugging Face Diffusers and have deep dive in inference frameworks like vLLM and SGLang, often diving into their codebases to understand how they improve performance.
PROFESSIONAL EXPERIENCE
Momo
AI Engineer • Mar 2022 - Present
Built AI systems: eKYC and GenAI products
Systems: 20M+ face vectors search engine, K8S deployment with vLLM & Triton
Patterned AI
AI Engineer • Core member
Built complete AI pipeline architecture for generating print-ready seamless patterns at scale
Serverless deployment: Modal infrastructure for efficient, scalable pattern generation
Cinnamon AI
AI Researcher • Feb 2019 - Mar 2022
Research focus: OCR, anime colorization, computer vision
EDUCATION
Ho Chi Minh City University of Technology
Honor Program in Electronics - Telecommunications Engineering Bachelor’s degree
2015 - 2019 • GPA: 8.05/10
PUBLICATIONS
-
Thien Do-Tieu et al. “Anime Sketch Colorization by Component-based Matching using Deep Appearance Features and Graph Representation”, International Conference on Pattern Recognition (ICPR), 2021, Pages 3154-3161
-
Trung D. Q. Dang, Thien Do-Tieu et al. “Correspondence Neural Network for Line Art Colorization”, SIGGRAPH ‘20: ACM SIGGRAPH 2020 Posters, August 2020, Article No.: 44, Pages 1–2
-
Hanh Phan-Xuan, Thien Do-Tieu et al. “Preserving Spatial Information to Enhance Performance of Image Forgery Classification”, International Conference on Advanced Technologies for Communications (ATC), 2019, Pages 50-55
-
Thuong Le-Tien, Thien Do-Tieu et al. “Image Forgery Detection: A low Computational-Cost and Effective Data-Driven Model”, International Journal of Machine Learning and Computing (IJMLC), 2019, Vol.9(2), Pages 181-188