Tan Ngo

Tan Ngo

AI Research Resident at Qualcomm AI Research
B.Sc. in Computer Science, Hanoi University of Science and Technology

Email · GitHub · LinkedIn · CV

I am an AI Research Resident at Qualcomm AI Research, working on 2D image, 3D content, and video generation — with an emphasis on distillation and quantization that make modern generative models deployable on edge devices. I graduated early from HUST in 3.5 years with a B.Sc. in Computer Science (GPA 3.69 / 4.0), finishing among the Top 9 graduating students and delivering the class graduation speech.

My research sits at the boundary between model design and systems: I want to understand AI end-to-end, from the mathematics of learning dynamics to the GPU architectures that make these models run at scale. I am currently applying to PhD programs to pursue this direction.

Generative models Video generation Diffusion distillation Multi-view / 3D Efficient inference Distributed training

News

Publications

3D-Grounded
Noise for Video
Track the Noise, Move the World: 3D-Grounded Motion-Consistent Noise for Controllable Video Generation
Long Van Vu*, Tan Van Ngo*, Animesh Karnewar, Amir Habibian, Binh-Son Hua, Hung H. Bui, Minh Hoai, Phong Ha Nguyen
Under review at NeurIPS 2026 In Submission
*Equal contribution

Research

Unifying Camera and Motion Control for Video Generation

Jan 2026 – Present

Qualcomm AI Research

  • Jointly controlling camera movement and object motion in generative video models.
  • Using depth estimation to inject 3D awareness, going beyond 2D-only conditioning.
  • Scaling training, inference, and evaluation across multi-node GPU systems with up to 300 GPUs.

Few-Step Text-to-Multiview Diffusion Model

Jun 2025 – Dec 2025

Qualcomm AI Research

  • Distilled a few-step MVDream-style model for higher-resolution multi-view generation.
  • Improved multi-view quality via training, data refinement, and model optimization.
  • Quantized the few-step model and supported deployment on Samsung Galaxy S24 phones.

Continual Learning for Object Detection with Generative Data Replay

Sep 2024 – Jan 2025
  • Applied continual learning to object detection to reduce catastrophic forgetting across sequential tasks.
  • Used diffusion-generated replay data to improve knowledge retention and training stability.

3D Scene Reconstruction with Neural Radiance Fields for 5G Antenna Inspection

Apr 2024 – Jun 2024

Viettel Digital Talent · Viettel Group

  • Reconstructed 5G BTS station scenes with NeRF and used the reconstructions to detect broadcast antenna tilt, reducing manual inspection cost.

Vision Transformers for Medical Endoscopy Segmentation

Oct 2023 – Aug 2024

Computer Vision Lab, BKAI, HUST

  • Vision Transformer methods for endoscopy image segmentation.
  • Used Masked Autoencoder (MAE) pretraining to improve segmentation classification.

Experience

AI Research Resident — Qualcomm

Jun 2025 – Present
  • Research on 2D image, 3D content, and video generation.
  • Distillation and quantization for efficient on-device inference (e.g., Samsung phones).
  • Developing video data generation methods for robot training.

Generative AI Engineer — ZenAI

Aug 2024 – Jun 2025
  • Optimized LLM inference and serving with vLLM and SGLang.
  • Accelerated inference via quantization, KV-cache optimization, and multi-GPU parallelism.
  • Reduced CPU–GPU communication overhead to improve serving throughput.

Viettel Digital Talent — Viettel Group

Apr 2024 – Jun 2024
  • Intensive training in data science, machine learning, and deep learning.
  • NeRF-based 3D reconstruction of 5G BTS scenes for antenna inspection.

Undergraduate Research Assistant — Computer Vision Lab, BKAI, HUST

Oct 2023 – Aug 2024
  • Vision Transformer methods for medical endoscopy image segmentation.
  • MAE pretraining to improve segmentation classification.

Education

Hanoi University of Science and Technology (HUST)

Aug 2021 – Apr 2025

B.Sc. in Computer Science · Hanoi, Vietnam

  • GPA 3.69 / 4.0 — graduated early in 3.5 years.
  • Top 9 Excellent Students of the graduating class.
  • Delivered the graduation speech on behalf of all graduating students.

Lam Son High School for the Gifted

Aug 2018 – May 2021

Mathematics Track · Thanh Hoa, Vietnam

Honors & Awards

Skills

Languages: Python, C / C++

Frameworks: PyTorch, distributed training and inference, multi-GPU / multi-node systems

Tools: Git, Linux, Docker, SLURM

Libraries: Diffusers, OpenCV, Pandas, NumPy, Matplotlib, scikit-learn