About Me
Who am I?
👋 Hi, I'm Le Viet Hai (Alex), an enthusiastic Data Science and Machine Learning practitioner with a passion for using AI to drive impactful solutions. With a strong foundation in Machine Learning acquired through courses at DeepLearning.AI, I am proficient in training models using PyTorch and applying Natural Language Processing (NLP) techniques.
🚀 My academic journey and internships have allowed me to work on diverse AI projects, including Time Series Data Analysis, Image Classification using ResNet/CNN, and NLP classification and prompting tasks. These experiences have not only honed my technical skills but also deepened my appreciation of AI's transformative potential.
🎯 Currently, I am focused on enhancing my skills to excel in upcoming internship opportunities. I'm dedicated to mastering algorithms on the LeetCode platform, understanding complex model architectures, and refining my problem-solving techniques to tackle real-world challenges.
🌟 My ultimate goal is to leverage my expertise in AI to create innovative solutions that improve lives. Let's connect and explore opportunities to collaborate in the ever-evolving world of technology!
Education
Singapore University of Technology and Design (SUTD)
BSc in Computer Science
Grade: 4.8 / 5.0
At SUTD, I was awarded the ASEAN Undergraduate Scholarship and named to the Honour List for academic excellence. I actively contribute to campus life as an executive member of the AI Singapore Student Club and have showcased projects at the SUTD Open House. I was also a winner of the 2024 SUTD What The Hack competition and currently participate in the Undergraduate Research Opportunities Programme (UROP) to gain hands-on research experience.
Tianjin University (TJU)
Short Summer Programme
I joined a cultural and academic exchange in Tianjin and Beijing, deepening my understanding of Chinese traditions and language. During the programme, my team won the Best Optimization Prize in the AI Car Path Optimization Challenge, jointly organized by SUTD and TJU.
Vietnam School of Science (VSSS)
Summer Research Program
Selected as one of the top 150 students nationwide, I participated in the 2022 Vietnam Summer School of Science themed "Stories of Life." This experience gave me valuable exposure to the research process, introduced me to inspiring mentors, and sparked my long-term interest in science and technology.
Nguyen Trai Specialized Senior High School for the Gifted
Mathematics Track
GPA: Grade 10 – 9.2 | Grade 11 – 9.2 | Grade 12 – 9.5 / 10.0
As a member of the Honors Mathematics Class and Vice Monitor of my cohort, I maintained consistent academic excellence and was awarded a Merit-based Scholarship for outstanding performance. I was also named to the Dean's List every semester.
Skills & Expertise
Programming Languages
AI/ML Frameworks & Web Technologies
Cloud Platforms & Infrastructure
Work Experience
AAAI Undergraduate Consortium Scholar
Association for the Advancement of Artificial Intelligence
AAAI Undergraduate Consortium
🔗 View ProjectKey Achievements
- Selected as an AAAI Undergraduate Consortium Scholar (10% acceptance), receiving professional mentorship in AI research and conference preparation.
- Investigated AI safety and alignment in quantized models, analyzing 4-bit QLoRA safety degradation and presenting findings at the Undergraduate Consortium Poster Session.
- Engaged in advanced research seminars and technical discussions focusing on the robustness and security of Large Language Model (LLM) systems.
GStar Bootcamp (Merit Graduate)
Key Achievements
- Completed the GSTAR Bootcamp with Merit and selected for a fully funded AI research programme (17% acceptance rate).
- Received advanced training from Google DeepMind, OpenAI, and Stanford experts on transformer architectures, pre-/post-training, AI safety evaluation, and agentic AI.
- Developed a lightweight delegation framework using a novel Confidence Token Routing method, achieving 85% accuracy on math_500—matching high-end Gemini performance at a lower cost.
Research Assistant UROP
Singapore University of Technology and Design
Context-Aware Interaction Lab, SUTD
🔗 View Research PaperKey Achievements
- Co-authored a CHI 2025 Late Breaking Work paper on evaluating Multimodal Large Language Models (MLLMs) in detecting joint attention in parent-child interactions.
- Conducted data collection and annotation of parent-child interaction videos in collaboration with speech-language pathologists.
- Apply Google Cloud ML infrastructure to run large-scale inference with Gemini models on video reasoning tasks.
AI for Problems Solving
Institute for Inforcom Research (I2R)
Agency for Science, Technology and Research (A*STAR)
Key Achievements
- Implemented advanced prompt engineering techniques with APIs such as GPT-3.5, LLaMA, and Mixtral to tag relevant knowledge components based on the Common Core State Standards for Mathematics.
- Developed and executed comprehensive training and evaluation frameworks for multiple Large Language Models (BERT, Roberta, Phi-3-mini, Qwen), optimizing performance for multiclass knowledge components classification tasks, achieving 95% accuracy within the top 5 classes using ASDiv data.
Research Assistant UROP
Singapore University of Technology and Design
ACSP Lab, SUTD
Key Achievements
- Designed a novel encoding method to minimize the sneak path effect by enforcing fewer memory cells with the Low-Resistance State (LRS).
- Develop a universal framework for 2D code design that imposes stronger constraints compared to existing literature, while maintaining the same level of complexity.
- Implemented modifications to the swapping technique to minimize the weight of each row and column in 2D binary codes, optimizing the overall performance and efficiency of the encoding scheme.
Student Researcher (Data Science: Deep Learning)
Key Achievements
- Engaged in an intensive 2-week Data Science 2023 Cohort focusing on Deep Learning, being recognized for exceptional contributions by MaSSP and the Vietnam Institute for Advanced Study in Mathematics (VIASM).
- Implemented an innovative approach using pre-trained models for word tokenization and embedding within a 2-layer Bidirectional LSTM Network.
- Achieved results with a 95.83% accuracy and an F1 score of 0.8684 on the ReINTEL dataset, showcasing the model's effectiveness in classifying false news articles.
Research Study Assistant
Key Achievements
- Engaged in analyzing electrocardiogram (ECG) data for a heart disease classification project, focusing on six common cardiac issues.
- Employed PyTorch library for developing and refining neural network models, gaining insights into the framework through experimentation with various architectures for time series data.
- Achieved notable results with an F1_mean of 0.9, utilizing a modified ResNet-18 architecture with an optimal model size of 33.43 MB.
Featured Projects
A showcase of my AI/ML projects, hackathon wins, and technical achievements. Click on any project card to view full details.
DripSeek
Top 4 out of 300+ participants with AI-powered browser plugin for real-time fashion detection and virtual try-on using AWS Rekognition and serverless APIs.
Towards Chain-of-Thought Reasoning for Video Understanding with Gemini: An Agentic Chain-of-Chain-of-Thoughts Framework
First-place AI research project developing Agentic Exploratory CoCoT framework with optimized prompts and video preprocessing for enhanced temporal comprehension using Gemini API.
Smart Health
Full-stack Android health tracking app with 15+ modules, OpenAI-powered medicine identification, and 31+ RESTful APIs built with Spring Boot and cloud infrastructure.
Deficiency Severity Classification
Runner-up maritime safety AI project using BERT-based consensus framework to predict deficiency severity levels with 60% accuracy, addressing subjectivity in maritime inspections.
Digital Fridge
Led the development of Digital Fridge, a 1st-place award-winning Flask app with Computer Vision and AI-powered recipe suggestions for automated food inventory management.