Writing
Research notes for curious people.
Plain-language tours through the papers, prototypes, and ideas behind my work.
When More Synthetic Speech Labels Stop Helping
A plain-language tour of our Interspeech 2026 paper on phonetic transcription, why synthetic labels help only up to a point, and what this means for accent-robust speech technology.
Speech AI Progress Is Not Evenly Distributed
A lay summary of Connecting the Dots, our draft on why automatic speech recognition improvements for mainstream English do not reliably transfer to minority English varieties.
Building Agricultural Search for Both Smartphones and Feature Phones
A lay summary of eKichabi v2, our CHI 2024 work on designing and scaling agricultural directory technology for nearly 10,000 rural Tanzanian enterprises.
Can AI Agents Estimate the Carbon Footprint of Electronics?
A plain-language breakdown of our Nature Electronics paper on multimodal AI agents that estimate electronics carbon footprints from public information.
Giving AI Agents Real Hardware to Learn From
A lay summary of Embedded Arena, our benchmark and hardware-in-the-loop harness for AI agents optimizing machine learning models on microcontrollers.
Making Product Sustainability Visible While You Shop
A plain-language summary of Living Sustainability, our IMWUT work on in-context interactive environmental impact communication for everyday product decisions.
How Many Holes Does a Graph Need?
A friendly explanation of our graph genus algorithm, why drawing graphs on surfaces matters, and how rotations helped compute the genus of the (3,12)-cage.
A Gentle Door Into Graphs on Surfaces
A lay summary of my UW honors thesis, which builds intuition for topological graph theory and the Practical Algorithm for Graph Embedding.