GPU Training Profiler
Real-time profiling for multi-GPU PyTorch training on Slurm: NVML telemetry, CUDA kernel stats, and step-phase breakdowns.
Software Engineer · ML & Systems
CS @ University of Florida. I build data pipelines, GPU tooling, and ML systems.
GitHubCS student at UF, into the layer where machine learning meets infrastructure. Previously interned at Google and JPMorgan Chase.
2026
Built production SQL + Python pipelines tracking GPU fleet data for ML program managers.
2025
Shipped FastAPI and React features that cut dataset retrieval time by 60%.
2025
Built a Gemini-powered multi-agent chatbot platform for HCI research.
Real-time profiling for multi-GPU PyTorch training on Slurm: NVML telemetry, CUDA kernel stats, and step-phase breakdowns.
Predicting a chaotic double pendulum: LSTM vs. an energy-conserving Hamiltonian Neural Network, trained on 3M+ simulated points.
PDF question-answering with Gemini, a Qdrant vector store, and Inngest-orchestrated ingest workflows.
Investment simulator with live prices, AI-summarized news, and LSTM forecasting.
FedAvg under attack from malicious clients.
PyTorch · Flower
Full-stack mutual fund prediction app.
Next.js · Flask
Java + C++ services over gRPC with payment deduplication.
Java · C++ · gRPC
Ranks teams by schedule-adjusted expected record.
Flask · MongoDB
A* and Dijkstra grid visualizer.
Python
Mobile robot simulation in ROS 2 and Gazebo.
ROS 2 · Gazebo
Open-source contribution to ColorStack's monorepo.
TypeScript
Concurrency lab from Bloomberg's BBIT series.
Python