Software engineer · Washington, D.C. area
James C. Scott
I build the platforms and tools other engineers rely on.
I've worked at Google, Microsoft, Capital One, Intel, the federal government's 18F, and a startup, on cloud platforms, developer tools, and the services behind them. Lately I work on AI tools for developers and on measuring how reliable they are.

Selected work
Google
An open-source developer platform
I set the technical direction for webstatus.dev, which shows which web features work in which browsers. It runs on Google Cloud with a public API. I led it through the security, privacy, and open-source reviews needed to launch it.
webstatus.dev ↗web-platform-tests
Test infrastructure the browser makers share
Apple, Google, Microsoft, and Mozilla all run web-platform-tests, a shared test suite. I'm on its infrastructure team: I help run its CI and results dashboard and review changes from engineers at several companies. I also wrote the proposals that added JavaScript's official test suite to it and linked its tests to web features.
wpt.fyi ↗Google
Measuring an AI tool's accuracy
wpt-gen is an open-source AI tool that writes browser tests. I defined a benchmark for its AI reviewer, oversaw the engineer who built the benchmark harness, and set up the pipeline that runs it. It checks whether the reviewer is right, consistent, and agrees with human reviewers. The team has used it to choose which model to run.
wpt-gen ↗Capital One
Distributed computing for data scientists
I led the engineers who added distributed computing (Amazon EMR) to the machine learning platform most teams there already used. Dozens of teams adopted it.
Coming soon · a series
A DGX Spark in my server room
Setting up an NVIDIA DGX Spark at home and measuring it: speed, accuracy, power, and what it costs compared with the cloud.
- 01Unboxing and headless setup
- 02Remote access
- 03First models and benchmarks
- 04Evaluating local models
- 05Power, noise, heat, and cost


