AI RESEARCH ENGINEER · HO CHI MINH CITY
From research ideas
to real-world AI.
I work at the intersection of computer vision, multimodal learning, and intelligent systems—bringing research into practical AI products.
01 — RESEARCH & SELECTED WORK
Bridging domains, building trust,
and making AI useful.
PUBLICATION · WACV 2025
Stratified Domain Adaptation: A Progressive Self-Training Approach for Scene Text Recognition
A progressive self-training approach that addresses the synthetic-to-real domain gap in scene text recognition, achieving state-of-the-art performance on benchmark evaluations.
SUMO-based microscopic traffic simulation for Ho Chi Minh City, combining OpenStreetMap roads with bus, car, and motorbike demand. Research on adaptive reinforcement learning for scalable traffic control under uncertainty.
VINUNIVERSITY · CAIR 2026 — Present
Fine-tuned Qwen2.5-VL and InternVL3 for expense management, merchant validation, and text extraction. Applied DPO and MPO during adaptation and led two collaborators on Vietnamese-language data collection and filtering.
MOMO VL 2025 — 2026
Developed four AI modules for passive liveness detection with screen-generated color sequences, targeting photo spoofing, 3D masks, deepfakes, and injection attacks. Contributed to MoMo Face Authentication achieving iBeta Level 2 certification.
FLASHMARK · MOMO 2024 — 2026
Developing an LLM-powered multi-agent workforce that combines tool use and workflow automation for business applications.
AGNTIC.AI 2026 — Present
A top-down 2D hand-pose pipeline combining an SSD hand detector with a Stacked Hourglass Network. Includes FreiHAND training and evaluation, plus a real-time single-hand camera demo.
View repository ↗HAND POSE ESTIMATION Public repository
A content-based image retrieval project using Faiss for indexing and similarity search. Compares VGG16, ResNet50, local binary patterns, and RGB histograms, with retrieval evaluation and a Streamlit demo.
View repository ↗CONTENT-BASED IMAGE RETRIEVAL Public repository
02 — EXPERIENCE
Research depth.
Engineering practice.
Aug 2026 — PresentINDUSTRY
LLM-powered multi-agent workforce, tool integration, and workflow automation for business applications.
May 2026 — PresentRESEARCH
Adaptive reinforcement learning for scalable traffic control under uncertainty. Built a Ho Chi Minh City traffic simulation, developed the bus demand pipeline, automated fixed-time traffic signals, calibrated car and bus parameters, and stabilized sublane motorbike behavior.
Supervisor: Dr. Duc Thinh Pham
Aug 2024 — Apr 2026INDUSTRY
Applied multimodal research and biometric security through MoMo VL and Flashmark eKYC. Managed two collaborators for Vietnamese-language data work and four collaborators for real-user and spoofing data collection.
Excellent Contribution Award · 2025
Apr 2024 — Jul 2024RESIDENCY
Viettel Digital Talent 2024. Developed an approach to non-intrusive speech quality assessment (NISQA) by integrating CMGAN speech enhancement.
Feb 2023 — Jul 2024RESEARCH
Led StrDA as first author, from methodology and experimental evaluation to manuscript preparation. Published at WACV 2025 (CORE A).
Supervisor: Dr. Thanh Duc Ngo
Jun 2023 — Aug 2023TRAINING
Built a 3D motion-capture system using E²FGVI video inpainting and FrankMocap pose estimation to align a mannequin with the scene.
03 — EDUCATION
NOV 2020 — MAY 2024
VNU-HCM · B.S. Computer Science
Honors Program
UNDERGRADUATE THESIS · 9.9 / 10Gradual Domain Adaptation in Scene Text Recognition
Supervised by Dr. Thanh Duc Ngo
AUG 2025 · RESEARCH TRAINING
Rencontres du Vietnam. Selected for interdisciplinary training in research methodology, academic integrity, critical thinking, and publishing. Led a Science-A-Thon team presenting on science, technology, and human resource development.
SEP 2017 — MAY 2020
Da Nang · Physics specialization. Grade 12 GPA: 9.6/10, first in class. First Prize in the Provincial Physics Competition (2020) and Second Prize in the Provincial Science and Engineering Fair (2019).
04 — TOOLKIT & RECOGNITION
PyTorch · TensorFlow · scikit-learn · Hugging Face
vLLM · SGLang · NVIDIA Triton Inference Server
AWS · Docker · Gradio · ComfyUI
SUMO
Vietnamese: Native · English: IELTS 6.5
Leadership: Class Chapter Secretary, Ho Chi Minh Communist Youth Union at UIT; Executive Committee Member, Green Season Journalism Club at Le Quy Don High School for the Gifted.
05 — LET’S CONNECT
For research collaborations and AI engineering opportunities.
khalee230702@gmail.com06 — AROUND THE WEB
Code, credentials, research, and everyday connections.