Software can be intentional.

Henry Burke

Software Engineer

About me

Henry Burke

doing cool stuff

I’m Henry, a software engineer who likes building small, finished things — tools for games I actually play, puzzles I wished existed, and, along the way, research on how large language models make sense of source code.

The throughline is care: type-checked puzzle grids, vision models that squint at game screenshots, and interfaces that respect people’s attention. Most of what I make starts as something I wanted for myself and ships once a stranger could enjoy it too.

Work

Projects

Burntbase — Instantly scan bases for 3-star attacks

Burntbase

A web and mobile app for Clash of Clans players. Upload a screenshot of an enemy base and an AI vision model identifies the layout and defenses, then surfaces YouTube videos of three-star attack strategies for that specific base.

Next.jsTypeScriptComputer VisionAWS S3
Crossword — The Mini daily crossword

Crossword

A SvelteKit mini crossword with an interactive grid, clue navigation, error checking, and a confetti finish. Puzzle data is defined in code so new grids are easy to ship.

SvelteKitTypeScriptSvelte 5

Writing

Research

ICSME 2023

Benchmarking Causal Study to Interpret Large Language Models for Source Code

Daniel Rodriguez-Cardenas, David N. Palacio, Dipin Khati, Henry Burke, Denys Poshyvanyk

Introduces Galeras, a causal benchmarking strategy for interpreting LLM performance on code completion, summarization, and commit generation — controlling for confounders that accuracy metrics miss.

arXivLLMsCausal Inference