AI Research Engineer

LY VINH THUAN

Specializing in multimodal architectures, scalable RAG & LLMOps. Building compact models that punch above their weight class.

01

ABOUT

I'm a final-year Data Science student at University of Science, HCMC, currently an AI Engineer at Zalo AI (VNG). I research and build multimodal AI systems, audio, vision, language, with a focus on making compact models that outperform much larger ones.

I compete solo in international computer vision challenges and publish across CVPRW, ICCVW. I care about inference efficiency, observability, and systems that actually work in production.

Multimodal AI RAG / LLMOps Computer Vision Audio-Language Models LoRA Fine-tuning PyTorch vLLM FastAPI
02

EDUCATION

Bachelor of Data Science Aug 2022 — 2026
University of Science, Ho Chi Minh, Vietnam

GPA: 3.6/4.0, 8.5/10

03

EXPERIENCE

AI Engineer Sep 2025 — Present
Zalo AI · VNG Corporation

Fast-tracked from Zalo Tech Fresher 2025 (1% acceptance).

AI Research Engineer Apr 2025 — Sep 2025
Kyanon Digital

Fast-tracked from Kfresher (6% acceptance).

Undergraduate Research Assistant Sep 2024 — Aug 2025
Aisia Lab · University of Science HCMC

Investigated multimodal sarcasm detection via text, image, and context fusion.

04

PUBLICATIONS

TinyGiantALM
Interspeech 2026
TinyGiantALM: A Compact Audio-Language Model for Intent-Aware Reasoning
Sole Author View →
FSM-Net
CVPRW · NTIRE 2026
FSM-Net: An Efficient Frequency-Spatial Network for Real-World Deblurring
Sole Author View →
TinyGiantVLM
ICCVW · AI City 2025
TinyGiantVLM: A Lightweight VLM for Spatial Reasoning under Resource Constraints
First Author View →
Dual-Path Enhancements in Event-Based Eye Tracking
CVPRW 2025
Dual-Path Enhancements in Event-Based Eye Tracking: Augmented Robustness and Adaptive Temporal Modeling
Second Author View →
Event-Based Eye Tracking
CVPRW 2025
Event-Based Eye Tracking
Co-Author View →
05

AWARDS

#2
NTIRE 2026 Efficient Real-World Deblurring
CVPR 2026 · Solo Competitor
Apr 2026
#4
Event-Based Eye Tracking Challenge
CVPR 2025
Jun 2025
#5
AI City Challenge Track 3 , Warehouse Spatial Intelligence
ICCV 2025 · NVidia
Oct 2025