Welcome to my GitHub profile! I'm Dung Vo Pham Tuan, currently pursuing my Master's degree in Computer Science with a specialization in Applied Data Science at Ho Chi Minh University of Technology (HCMUT).

I have a strong background in mathematics and a passion for Artificial Intelligence, with experience in Data Science, Natural Language Processing, and Computer Vision. My goal is to become a Professional Machine Learning Scientist, continuously improving my expertise and contributing to cutting-edge AI research.

• 🔭 I'm currently working on Text-based Person Re-identification and Knowledge Distillation for LLMs

• 🌱 I'm currently learning MLOps, Vector Databases, and Large Language Model Optimization

• 👯 I'm looking to collaborate on AI Research Projects in Computer Vision and NLP

• 📫 How to reach me: vophamtuandung05hv@gmail.com

• 📄 Check out my portfolio website for more details

Dien Toan Group | Jul 2024 - Apr 2025 | Tan Binh District, HCM City

• Pretrained a multilingual vision-language backbone (Vietnamese/English/Chinese) for Text-based Person Re-identification on a large-scale dataset (36 million image-text pairs) using 4 NVIDIA A100 GPUs
• Pretrained/Fine-tuned models for Text-based Person Re-identification using custom PyTorch implementation without relying on high-level training frameworks

• Extended the original English pretraining datasets by adding Chinese and Vietnamese captioning annotations, demonstrating that multilingual pretraining improves zero-shot retrieval performance by over 1.2% Rank-1
• Leveraged advanced LLMs and MLLMs with optimized inference frameworks (vLLM, lmdeploy, SGLang) to accelerate data augmentation, enhancing dataset diversity and quality for improved model performance

Dien Toan Group | Oct 2023 - Jun 2024 | Tan Binh District, HCM City

• Proposed shifting the company's focus from fixed-attribute person re-identification to Vietnamese Text-based Person Re-identification, a more practical solution for the local context

• Constructed the first Vietnamese pretraining and benchmark datasets for this task, significantly enhancing fine-tuning efficiency and model generalization
• Developed a Vietnamese Vision-Language backbone based on the ALBEF architecture, integrating SOTA Vietnamese language models such as PhoBERT (VinAI Research) and ViDeBERTa (FSOFT AI)
• Improved the image encoder stream using architectures from the HAP and SOLIDER frameworks, leveraging Vision Transformer/Swin Transformer models pre-trained on human-centric surveillance datasets

Dien Toan Group | Jun 2023 - Jul 2023 | Tan Binh District, HCM City

• Conducted comprehensive survey and analysis of research papers on Transformer-based architectures for Object Detection and Multiple Object Tracking
• Implemented and comprehended the underlying mechanisms of Trackformers (Facebook AI Research, CVPR 2023) for tracking pedestrians and vehicles at the campus of Ho Chi Minh University of Technology
• Preprocessed realistic surveillance video data using OpenCV and FFmpeg for frame extraction, noise reduction, and format standardization
• Evaluated multiple efficient data annotation tools for creating high-quality tracking training datasets
• Optimized Trackformers by modifying architecture/loss to extend from single-class (human-only) to multi-class tracking and mitigate class imbalance, where pedestrian instances significantly outnumbered vehicles in the dataset
• Demonstrated real-time system performance in multiple real-world environments, including a technical presentation at Ho Chi Minh City University of Technology (HCMUT)

Data Science Lab, CSE Faculty, HCMUT | Aug 2023 - Feb 2025 | District 10, HCM City

• Conducted academic research on Text-based Person Re-identification

• Served as the sole researcher responsible for the entire project, developing a state-of-the-art model that achieved a 2.8% Rank-1 accuracy improvement on benchmark datasets over recent SOTA models
• The resulting models became the highlighted AI products of the lab and serves as a benchmark for future research
• Due to the sensitivity of surveillance camera data and commercial constraints, the work is being developed as an internal proprietary product rather than submitted for publication
• Developed and publicly disseminated technical documentation detailing key innovations in training pipelines, ablation studies, and multi-modal result visualizations to benefit the broader technical community

Master's Degree in Computer Science | Jan 2024 - Present

• Specializing in Applied Data Science
• Current GPA: 8.48/10 (24/60 credits)

Bachelor's Degree in Computer Science | Aug 2020 - Nov 2024

• Honors Degree with dual specializations in Image Processing & Computer Vision and Applied Artificial Intelligence
• GPA: 8.69/10 (3.8/4) - Thesis Score: 9.7/10 (AI Research)

High School Diploma | Aug 2017 - Jul 2020

• Specialized in Mathematics with a GPA of 9.4/10
• Direct Admission to University due to Third Prize, Vietnam Mathematical Olympiad 2020

Individual Project | Apr 2025

• Implemented an open-source knowledge distillation framework (GitHub repo) to transfer structured reasoning from GPT-4o to a mini-LLM (Qwen2.5 Coder 1.5B Instruct) for Coding Multi-Choice Coding Question Answering
• Generated a YAML-based reasoning dataset from a subset curated samples from CodeMMLU using GPT-4o, leveraging OpenAI SDK for the data synthesis pipeline
• Designed a structured reasoning framework (understanding question → analysis choices → reasoning → conclusion → answer) that mirrors how CS students and researchers systematically approach problems
• Fine-tuned the model using parameter-efficient techniques (LoRA, Lion optimizer) with advanced optimization strategies (gradient checkpointing, mixed precision training,...) for memory efficiency
• Created a 4-bit quantized interactive live demo on Hugging Face Spaces (live demo) for evaluation and demonstration
• Developed comprehensive training analytics with WandB integration for experiment tracking, including prompt monitoring, token distribution analysis, and quality metrics
• Structured the repository with modular components and comprehensive documentation including setup guides, architecture diagrams, and clearly explanations for reproducibility

Individual Project | Mar 2025

• Designed and developed a full-stack RAG system using FastAPI, Weaviate, and OpenAI SDK, with a self-hosted vector database for data control and privacy (GitHub repo)
• Implemented a containerized microservices architecture with Docker Compose local development environment
• Extended deployment to Google Kubernetes Engine (GKE) using Terraform for infrastructure provisioning and Kubernetes manifests for orchestration
• Implemented CI/CD pipelines with GitHub Actions for testing individual components, API testing, building Docker images, and pushing them to Docker Registry
• Developed a text processing pipeline for efficient document chunking and optimized vector search performance
• Created a user-friendly web interface for document upload, search, and question answering using Streamlit

Individual Project | Oct 2023

• Worked with the Human Action Recognition benchmark from a Kaggle contest
• Proposed a modern approach by fine-tuning the vision-language model CLIP for an open-vocabulary detection task, replacing traditional fixed-category classification
• Reimplemented the training/inference pipeline, Trainer class with similar functionalities like Transformers library, just by PyTorch without using external training frameworks
• Integrated experiment tracking, data versioning, and model registry to streamline training experiments using MLflow
• Implemented a CI/CD pipeline with GitHub Actions to automate Docker image builds and deployments
• Deployed the model API using FastAPI and managed scalable infrastructure on Google Kubernetes Engine (GKE) using Terraform
• Created a user-friendly web interface for real-time human activity recognition using Streamlit

Team Project - Team Lead | Dec 2022

• Led a team of four, responsible for coordinating overall project development and managing the project timeline
• Specialized in designing algorithmic solutions and creating mathematical models for the puzzle-solving agents
• Developed AI agents to solve Water Sort, Bloxorz, and Vietnamese Reversi games
• Engineered intelligent agents using A*, Minimax, Genetic Algorithms, and Deep Q-Learning (reinforcement learning) to optimize game strategy and performance
• Developed user interfaces and visualizations for game environments using PyGame, OpenGL and Cocos3d, while implementing the AI agents' logic and training with PyTorch for Deep Q-Learning

• Honors Degree Graduate with Dual AI Specializations and GPA 8.69/10 (Nov 2024)
• Student of Five Merits at Vietnam National University level and Ho Chi Minh City level (Nov 2024)
• Third Prize, Faculty Thesis Poster Competition For Talent Students (Top 3 Thesis) (May 2024)
• University Incentive Scholarship for Outstanding Students (Sep 2023)
• Odon Vallet Scholarship For Outstanding Vietnamese Students (Sep 2020)
• Third Prize, Vietnam Mathematical Olympiad (VMO) (Jan 2020)
• Consolation Prize, Vietnam Mathematical Olympiad (VMO) (Jan 2019)
• Gold Medal with Top 5, April 30th Mathematics Olympiad for Gifted Students in Southern Vietnam (Mar 2019)
• Gold Medal with Top 1, April 30th Mathematics Olympiad for Gifted Students in Southern Vietnam (Mar 2018)

• Critical Thinking
• Communication Skills
• Problem-Solving
• Time Management
• Leadership
• Quick Learning
• Languages: English (Professional working proficiency, TOEIC 760/990)

• DevOps Professional Certificate — PagerDuty & LinkedIn (Mar 2025)
• Building Cloud Computing Solutions at Scale Specialization — Duke University & Coursera (Aug 2024)

• Large Language Model Operations (LLMLOps) Specialization — Duke University & Coursera (Jun 2024)
• Machine Learning Operations (MLOps) Specialization — Duke University & Coursera (Jun 2024)
• Machine Learning Engineering for Production (MLOps) Specialization — DeepLearning.AI (Feb 2024)
• Vector Databases Professional Certificate — Weaviate (Jul 2024)

• Building Real-Time Video AI Applications — NVIDIA (Aug 2024)
• Generative AI with Diffusion Models — NVIDIA (Aug 2024)

• Generative AI for Data Scientists Specialization — IBM (May 2024)
• Machine Learning Professional Certificate — IBM (May 2024)
• Advances In Natural Language Processing Specialization — VietAI & New Turing Institute (Mar 2024)
• Large Language Models Professional Certificate — Databricks (Oct 2023)
• Generative Adversarial Networks (GANs) Specialization — DeepLearning.AI (Jul 2023)
• AI Engineering Professional Certificate — IBM (Jul 2023)
• Natural Language Processing Specialization — DeepLearning.AI (Jul 2023)
• Deep Learning Specialization — DeepLearning.AI (Jun 2023)
• TensorFlow Developer Professional Certificate — DeepLearning.AI (Jun 2023)

Thanks for visiting my profile! Feel free to reach out for collaborations.