Research → product

Useful systems for the moments that matter.

I turn work on software testing and trustworthy AI into tools that help people build, teach, and learn with better evidence.

OneAI · AI education

OneAI

A teacher-first workspace for creating assessments, guiding student reasoning, and turning rubric decisions into feedback that students can act on.

Watch the walkthrough
01

Coach the next step

Students get a hint or a check on their reasoning while the work is still theirs.

02

Build a complete paper

Generate a structured test from a prompt, then export it in the format your class uses.

03

Keep teachers in control

Rubrics make scoring consistent; teacher review keeps the final judgement human.

04

Explain every gap

Per-student error reports show the missed idea and a concrete route to practice it.

Product walkthrough

From first question to useful feedback.

One flow for students, teachers, and the evidence between them.

OneAI combines guided student chat, paper building, rubric-based scoring, and student-level error reports in one workspace.

Pacific Studio · selected products

Three products. One Pacific Studio thesis.

Pacific Studio’s tools connect the same questions I work on: how agents build software, how we prove it, and how people understand the research behind it.

01 · BuildReleased

OpenASE

Issue-driven orchestration for the full software lifecycle, with Claude Code, Codex, and Gemini CLI working across design, development, testing, and deployment.

View on GitHub
02 · ProveReleased

VeriFormal

AI-assisted formal verification that turns plain code into machine-checked proof goals, with an offline workflow for C, Java, Dafny, and Rust.

Start verifying
03 · UnderstandBeta

KeyLight

An interactive paper reader that combines layout-aware parsing, sentence-level highlights, bilingual reading, and answers that point back to the source.

Try the beta
View Pacific Studio

For engineering teams

Make passing tests mean more.

My research on UTBoost and SWE-Bench reliability becomes practical test augmentation for teams shipping coding agents. Stronger tests expose semantic regressions before a green check reaches production.

Work together

Have a hard-to-measure workflow?

I work with research teams, product groups, and educators to turn a fuzzy quality question into a testable workflow.

Start a conversation