About
I believe the next frontier in AI is not reasoning ability but behavioral fidelity:
the gap between what AI can compute and how it actually acts in the physical world.
My work at Stanford HAI sits at this intersection, building systems that can perceive,
measure, and reason quantitatively about physical environments, grounded in how humans
actually work and learn.
Before Stanford I spent five years shipping production AI infrastructure at
Airbnb and Tableau, which taught me that the hardest
problems in AI are rarely algorithmic. They are about closing the gap between what a
system knows and what it can reliably do.
React / TypeScript
Python
BERT / NLP
Computer Vision
On-device LLM
Serverless / AWS
CI/CD
Data Visualization
Systems Architecture
HCI
Research & Projects
QuantiPhy Challenge
NeurIPS 2026
Platform administrator and organizer for the first large-scale benchmark on
quantitative physical reasoning for vision-language models. Co-organized with
Prof. Fei-Fei Li
and
Prof. Ehsan Adeli
at Stanford HAI.
ConfidentLM
Course Project
On-device LLM infrastructure with local inference, memory-efficient model execution,
and integrated hallucination detection and confidence verification pipelines.
Supervised by
Prof. Andrew Ng.
visx
Core maintainer of Airbnb's open-source React data visualization library.
20k+ GitHub stars, widely adopted across the industry. Built at Airbnb,
maintained as an ongoing open-source contribution.
Patternize.io
An e-book explaining data structures and algorithms through user-friendly D3
visualizations. ~500 stars on GitHub.
Education
M.S. in Human-Computer Interaction
Co-organized QuantiPhy Challenge (NeurIPS 2026) with Prof. Fei-Fei Li ·
Supervised by Prof. Nick Haber, Autonomous Agents Lab
B.A. in Computer Science (HCI + AI)
Industry Experience
- Led product design and engineering for a 20-person team; achieved product–market fit and scaled to millions in ARR, securing $20M+ in funding.
- Led development for cross-platform applications spanning web, mobile, and interactive platforms, leveraging modern frameworks, backend systems, and computer vision technologies.
- Executed AI-driven media campaigns, boosting ROI from 89.8% to 217.4%.
ReactTypeScriptPythonAWS
ServerlessComputer VisionCI/CD
- Built customer agent tools and chatbot features, including reply templates that improved ticket resolution time.
- Rebuilt support classification with a BERT-based model, enabling the chatbot to resolve more issues autonomously.
- Automated customer support notifications, reducing related human support costs by $24M.
- Maintained
airbnb/visx (>20k GitHub stars), adding support for React 18/19.
- 2021 Hackathon Winner: developed a 3D Earth visualization showing travel history on Airbnb.
React 18/19TypeScriptBERT
NLPD3 / visxPython
- Built Tableau Public, enabling millions of users to analyze and publish data visualizations online; feature lead for hashtags, search autocomplete, and image/thumbnail generation with serverless architecture.
- Refactored Tableau Online to TypeScript, React, and Redux; set up CI/CD pipelines, accelerating feature delivery and improving developer experience.
- Delivered talks at Tableau Conference 2017, React Conference 2019, and AWS Re:Invent 2019.
TypeScriptReactRedux
ServerlessAWSData VizCI/CD
Talks
AWS Re:Invent 2019
Serverless data architecture for large-scale visualization platforms · Las Vegas, NV
React Conference 2019
Data visualization at scale with React and D3
Tableau Conference 2017
Building extensible data platforms for public audiences
