AI Developer. Building trust and verification infrastructure for autonomous agents.
M.S. Artificial Intelligence, Worcester Polytechnic Institute. Working on the layer between what agents can retrieve and what is actually true.
Small, focused tools. Each one exists because I needed it and it didn't exist.
| a11y-judgment | Accessibility review for the ~60% of WCAG criteria automated tools can't check, the ones that need a judgment call about meaning. |
| before-you-build | Checks how crowded an idea already is on GitHub before you spend a weekend rebuilding it. |
| gh-trend-search | GitHub trending search that handles the Search API's undocumented rate limits and query quirks, including the one where grouping parentheses silently return zero results. |
- Enhanced Large Language Models, IEEE ICICIT 2024. Context understanding and parameter-efficient fine-tuning.
- InternEase: Creating Pathways to Professional Success, ICDSMLA 2023.
- CHIPNet: Coarse-to-Fine Hierarchical Inference for Precise Corner Detection on Chips using Neural Networks, M.S. thesis. A three-stage hierarchical deep neural network for precisely locating corners in semiconductor chip images, reaching 99.06% patch detection accuracy. Advised by Prof. Ziming Zhang, in collaboration with Teradyne.
The public web is the published subset of the world's knowledge, not the whole of it. A great deal of true, valuable, verifiable knowledge is never written down anywhere, so agents can't retrieve it, and usually don't say so. They answer anyway.
Most of my current work sits somewhere near that problem.