Machine Learning  ·  Software Engineering  ·  Medical Imaging  ·  Generative AI

I am a Computer Science master's student at the University of Southern California, building software for machine learning and medical imaging. My work includes evidence-grounded LLM agents for training diagnostics, diffusion models for MRI translation, and computer vision systems.

• Interested in reliable ML systems, generative AI, and healthcare applications
• Based in Los Angeles and open to Summer 2027 Machine Learning / Software Engineering internships

Built a Gemini-powered agent with a custom Python tool-calling loop to inspect PyTorch source code, experiment configurations, and training telemetry. It produces evidence-linked diagnoses, validates targeted repair proposals, and checks recovery through bounded training reruns, with support for excessive learning rates, missing optimizer steps, and healthy-run abstention.

Current work extends training telemetry and optimizer-parameter diagnostics to a ResNet18 / Camelyon17-WILDS medical-image classification baseline.

Python PyTorch LLM Agents Tool Calling Pydantic ML Systems

Developed a PyTorch conditional diffusion model for bidirectional T1/T2 MRI translation, combining anatomy-consistent structural guidance, cross-attention, classifier-free guidance, and a composite MSE + L1 + SSIM objective. Achieved up to 0.90 SSIM, 27.5 dB PSNR, and 0.986 NCC, with an open-source training and inference pipeline.

Python PyTorch Diffusion Models Medical Imaging Computer Vision

Adapted OnePose++ into a multi-sensor pipeline for estimating an object's 3D position and rotation. Built data-collection and preprocessing tools, a Python backend, and Vue.js / JavaScript visualizations for real-time pose tracking.

Python Computer Vision Pose Estimation Sensor Fusion Vue.js

• CMCD: Conditional Diffusion Model for Medical Image Modality Translation — First author; ICONIP 2026 (International Conference on Neural Information Processing). Code
• S. Fan, Z. Lu, Z. Yan, and L. Hu. “Interactions of three berberine mid-chain fatty acid salts with bovine serum albumin (BSA): Spectroscopic analysis and molecular docking.” International Journal of Biological Macromolecules, 274(Pt 2), 133370, 2024.

Languages: Python, C++, C#, Java, JavaScript, SQL
Machine Learning & Vision: PyTorch, TensorFlow, Diffusion Models, Transformers, OpenCV, OpenPose
LLM & ML Systems: Gemini API, Tool Calling, Training Telemetry, Pydantic, pytest
Tools & Platforms: Git, Linux, Weights & Biases, Unity, Vue.js, Raspberry Pi, Arduino