I’m a Data Science student at UC San Diego, passionate about machine learning, distributed systems, and high-performance computing.
My work range from optimizing LLM inference systems on Raspberry Pi clusters to designing data pipelines and building real-time ML models for cloud deployment.
Languages: Python, Java, SQL, PostgreSQL, HTML, CSS, JavaScript/TypeScript, C++, R, Bash, MATLAB, Ansible
Frameworks/Libraries: PyTorch, TensorFlow, scikit-learn, Pandas, NumPy, Matplotlib, Plotly, OpenCV, Flask
Tools/Platforms: Git, RESTful APIs, Docker, Kubernetes, CUDA, VSCode, Jupyter, Linux, Azure, AWS, Slurm, MPI, CI/CD
Focus Areas: Machine Learning, Cloud Computing, HPC, Distributed & Parallel Systems, Data Engineering
- Exploring CUDA-based optimization for ML workloads
- Learning advanced backend & cloud orchestration tools
IEEE Supercomputing (UCSD) — Officer
- Benchmarked Stable Diffusion XL with MLPerf across multi-GPU and HPC clusters.
- Led sub-team for Distributed LLaMA, placing 2nd nationally in the D-LLaMA. Competition Results →
C++, CUDA, Linux, Docker
Developed a modular GPU-accelerated image processing system with separate CUDA kernels for grayscale, box blur, and Sobel edge detection.
Designed a clean multi-file structure with a custom Makefile build system, achieving efficient parallel performance across multiple RTX 2080 Ti GPUs.
Currently extending the project with Gaussian blur and benchmarking utilities for CUDA performance profiling.
Python, Pandas, scikit-learn, Matplotlib, Seaborn
Analyzed match data from professional League of Legends games to explore how gold distribution and role dynamics affect player performance.
Performed fairness and error analysis to evaluate model consistency across different game contexts.
Python, PyTorch, Raspberry Pi Zero
Built and deployed an LSTM model for real-time sentiment analysis on Raspberry Pi Zero.
Optimized hyperparameters via Bayesian search and achieved 97% validation accuracy, earning 2nd place at IEEE competition.
Python, Matplotlib
Implemented gradient-based algorithms (GD, AdaGrad, RMSProp, AdaDelta, Adam) from scratch.
Visualized optimization paths over complex surfaces using Matplotlib animations.
Python, scikit-learn
Developed a facial landmark–based detection pipeline with visual evaluation and cross-validated metrics.
Python, TensorFlow, CUDA
Researched preprocessing methods and used CUDA acceleration to achieve 95% validation accuracy under limited compute.
Python, LSTM
Created a 2D physics engine to generate motion data and trained an LSTM for trajectory prediction.
Awarded Bronze at KSEF.
Data Science Intern @ Cloocus (Microsoft Azure Partner) – Kuala Lumpur, Malaysia
Jul 2025 – Sep 2025
- Built retrieval-augmented generation (RAG) chatbots and standardized SQL pipelines for enterprise data.
- Integrated data ingestion between Azure Blob, SQL, and REST APIs for real-time dashboards.
- Supported secure ML deployments for government clients via containerized Azure infrastructure.