Do Tri Nhan

Do Tri Nhan

AI Engineer in Speech and Signal Processing

About Me

I am Do Tri Nhan, an AI Engineer specializing in Speech and Signal Processing, currently working as Senior AI Engineer, Conversational Speech Technology at Vinfast Automotive AI Development Institute - Vingroup JSC, Ho Chi Minh City, Vietnam.

Download CV (Short)

Research Interests

My work spans the full speech AI stack — from core research to production deployment:

  • Speech Synthesis (TTS): Flow Matching, FastSpeech2, Tacotron2, GlowTTS, voice cloning & conversion, zero-shot TTS, GRPO-based reinforcement fine-tuning, diffusion step distillation
  • Speaker Verification & Diarization: EEND-VC, streaming diarization, voice biometrics, anti-spoofing
  • Speech Self-Supervised Learning (SSL): WavLM, representation learning for Vietnamese speech
  • Text Normalization & Phonology: Vietnamese NLP, grapheme-to-phoneme, hybrid rule-LLM pipelines
  • MLOps: TensorRT, Triton Inference Server, TorchServe, Docker, FastAPI, RabbitMQ

Background

I completed my Bachelor of Science in Computer Science (Advanced Program, APCS K17) at VNUHCM – University of Science in 2021, graduating with a GPA of 3.83/4.0.

During my undergraduate years, I was a Research Assistant at the AI Lab and Robotics & IoT Lab at USSH, where I built a 19 GB Vietnamese speech corpus and researched text normalization and phonology. I later interned remotely at the Speech and Language Lab, Nanyang Technological University (NTU), working on sound event detection in urban environments (project with ST Engineering).

After graduating, I spent over 5 years within Vingroup JSC — starting as an Engineering Resident at VinAI Research Institute, then growing through AI Engineer and Middle AI Engineer roles at Vingroup Big Data Institute (working on TTS, speaker diarization, voice biometrics, and AI agents), followed by a leadership role as AI Team Lead at VinSmart Future. Since June 2026, I have been a Senior AI Engineer at Vinfast Automotive AI Development Institute - Vingroup JSC, focused on Conversational Speech Technology for next-generation in-vehicle AI. In this role, I am in charge of the TTS core module — building an automated annotation and GRPO training framework and applying diffusion step distillation to improve inference speed, reduce hallucinations, code-switching pronunciation, and natural pausing.

Open-Source Contributions

I maintain two widely used Vietnamese NLP packages on PyPI:

  • vinorm – Vietnamese Text Normalization (34,933+ downloads / 6 months)
  • viphoneme – Vietnamese Grapheme-to-IPA Phonetization (4,195+ downloads / 6 months)

Selected Awards

  • Outstanding Employee of the Year – VinBigdata (2023)
  • 1st Place, VLSP TTS Emotional Speech Synthesis (2022)
  • 1st Prize, Science-A-Thon, 9th Vietnamese Summer School of Science (2022)
  • Outstanding Student Award in AI, Ho Chi Minh City (2021)
  • Excellent Thesis – Science and Technology Student Award (2021)

Contact

Feel free to reach out at dotrinhan99@gmail.com or connect via LinkedIn and Google Scholar.