studyap.org
StudyAP
Learning toolsAI practice problems, study plans, and grading for AP students.
Next.jsTursoDrizzle
Seraph Yang
Selected work
From an idea to something you can use.
helius.vercel.app
Helius
Web applicationA personalized cold email platform with an AI assistant.
Next.jsTailwindSupabase
Alittlecuriosity
goesalongway.
Learning from unlabeled videoLatent Action World Models
Stanford CS224RLearning action representations from unlabeled video for robot planning and control.
Implemented a PyTorch pipeline and compared latent-action regularization on SSv2 and DROID.
PyTorchV-JEPA2World models
Read paper (PDF) Experience
Software, research, and a few questions that took me somewhere unexpected.
View full résumé
Undergraduate Researcher
Researching latent reasoning and continuous chain-of-thought models, including a framework for weighting explicit and latent tokens.
Research Intern
Labeled 14,000+ X-rays with RadGraph and trained a ResNet-101 model for respiratory disease diagnosis.
ISS Research Lab
2024–25Student Researcher
Developed Arduino/C++ automation for experiments studying phage interactions with S. aureus in microgravity.
Life lately
A few moments worth keeping.
A little about me
Pick an outfit
I'm studying CS and Math at Stanford, drawn to AI research, startups, and the process of turning an idea into something useful.
I also play golf and squash, go climbing, and take long walks. I like well-made watches, thoughtful everyday objects, and having things to be curious about outside of code.
Languages
Python, C / C++, Java
Web
React, Next.js, Turso, Supabase
Machine learning
PyTorch, TensorFlow, NumPy, pandas