Stanford University · Class of 2027
Computer Science and Management Science & Engineering. I work on AI for healthcare, robotics, and markets.
I'm an undergraduate at Stanford studying Computer Science (AI track) and Management Science & Engineering, building at the intersection of machine learning and the physical world.
My work spans robotics world models, medical imaging, clinical NLP, and tools for wildfire safety. Outside of research I analyze equities for Stanford's student-run endowment fund, and spend time on CAD, woodworking, and design. I grew up in Hemet, California.
World models that help robots anticipate how objects move and interact during manipulation tasks like grasping and stacking. Adapts the JEPA architecture over object tokens on the LIBERO benchmark, using a slot-attention tokenizer and an action-conditioned transformer predictor; extending the work toward publication.
An agentic assistant that answers natural-language questions over robot deployment logs and telemetry. Built with LangGraph for orchestration, tool use, and memory, routing between structured metric queries and retrieval over unstructured incident reports. Grounds answers with a FAISS vector store and returns source provenance to reduce hallucination, with Langfuse tracing and an evaluation harness measuring accuracy and latency.
A retrieval-augmented research pipeline that drafts and qualifies investment theses against source filings. Combines hybrid BM25 and embedding retrieval, LLM-based synthesis, and automated citation verification to ground every claim against source text. Integrated LLM APIs end to end with programmatic prompt orchestration and structured output parsing.
End-to-end deep learning pipeline detecting ACL tears from knee MRI scans on the Stanford MRNet dataset. Fine-tuned ResNet18 across sagittal, coronal, and axial views with a learned logistic combiner, exceeding a PCA + logistic regression baseline by a substantial margin.
NLP pipeline converting doctor-patient dialogue into structured four-section SOAP notes on the MTS-Dialog corpus. Stage one extracts clinical entities and routes them to the correct section; stage two generates clinical prose with a DSPy-based LLM agent, evaluated against published baselines.
An evacuation and wildfire-risk dashboard for California, mapping live fire perimeters, evacuation zones, and area risk to surface clear guidance during fast-moving events. A personal project, shaped by being directly affected by the 2022 Fairview Fire in Hemet, CA.
A site that forecasts surf conditions and flags safety hazards — swell, wind, tide, and rip-current risk — to help surfers decide when and where it is safe to paddle out.
Charles R. Blyth Investment Fund · Stanford, CA
Analyst for Stanford's student-run endowment fund supporting financial aid, covering technology and small-cap names across AI, data centers, and healthcare. Run DCF and comparable-company valuation alongside qualitative diligence on business quality, downside risk, and management.
Freeman Spogli Institute · Stanford, CA
Studied adoption of Kaigo robotics in Japanese long-term care facilities, modeling assistive-technology deployment against patient outcomes across 200+ facility-year observations. Built SQL ETL pipelines and mixed-effects regression models in R, improving accuracy 12% over baseline.
California Ear Nose & Throat · Hemet, CA
Supported patient intake, clinical documentation, and exam-room workflow; prepared clinical instruments and assisted with rhinoscopy procedures. Engineered a full-stack patient management system using React and Node.js.
Architecture, CAD, and robotics.
Reach me here.