Daniel Lee
Stanford CS & AI. Building the future of intelligent systems.
About Me
Computer Science and AI student at Stanford. I work on research systems and build tools that advance how we understand and interact with technology.
You can reach me at: leedan [at] stanford [dot] edu
Experience
- Undergraduate Student ResearcherStanford Artificial Intelligence Laboratory (SAIL) — Stanford, CAJun 2025 – PresentExploring AI agents for research applications (Zou Group).
- Software Engineer InternModern Nose Clinic — Salem, OR (Hybrid)May 2025 – PresentBuilding tools to streamline prior authorization and patient price estimation for a specialty ENT clinic.
- Machine Learning ResearcherStanford University Department of Computer Science — Stanford, CAJan 2025 – May 2025Developed entropy-based system for detecting LLM hallucinations with high accuracy, leveraging information-theoretic features.
- Bioinformatics InternIcahn School of Medicine at Mount Sinai — New York, NYOct 2022 – Jan 2025Built scalable Hi-C analysis pipeline for characterizing chromatin architecture in addiction-related pathways.
- Research AssistantMassachusetts General Hospital — RemoteMar 2021 – Mar 2022Built Java-based simulation modeling embryonic growth and phylodynamics across species.
Papers
ShED-HD: A Shannon Entropy Distribution Framework for Lightweight Hallucination Detection on Edge Devices
A new way to detect LLM hallucinations efficiently, even on edge devices.
How Our Cells Become Our Selves: The Cellular Phylodynamic Examination of Growth and Development
How a single cell becomes a complex organism, using math to reveal the hidden rules of growth.
Projects
Scopia Research
Full-stack academic research platform processing 2.8M+ papers with 94% search relevance accuracy. Engineered dual-mode search combining vector similarity and Retrieval-Augmented Generation (RAG), reducing query latency by 65%. Implemented asynchronous Python backend with Redis caching, cutting infrastructure costs by 40%.
Cellular Phylodynamic Growth Simulation
Java simulation analyzing embryonic growth and phylodynamic patterns across 13 species, achieving 89.9% accuracy in cellular growth modeling. Implemented Universal Growth Equation achieving R² > 0.99 fit for developmental trajectories in 6 species. Published third author in BioRxiv on mathematical framework analyzing 2 trillion cells across development.
ShED-HD: Shannon Entropy Distribution for Hallucination Detection
Entropy-based system for detecting LLM hallucinations with high accuracy, leveraging information-theoretic features and scalable training methods. Built for edge devices and published in collaboration with Stanford CS.