Selected work · Open source · Collaboration

I am Shurong Cao, a PhD researcher in Electronic Engineering at The Chinese University of Hong Kong, following my bachelor's education at Nanjing University.

I am drawn to difficult questions that do not stay neatly inside one field. My work is an evolving mix of research, engineering, experimentation, and open collaboration. Frontier AI is one space I am actively exploring: how advanced systems reason and fail, how we should evaluate them, and how they can become useful in real scientific and engineering work.

I care about turning promising ideas into things other people can inspect: runnable software, explicit tests, reproducible evidence, public releases, and honest records of what remains uncertain.

• The Chinese University of Hong Kong — PhD researcher in Electronic Engineering.
• Nanjing University — bachelor's degree.

This is the foundation of my work, not its boundary.

These are current questions rather than permanent labels:

• How can frontier AI systems support scientific and engineering reasoning without hiding uncertainty, limitations, or failure?
• What makes an experiment, benchmark, or software result genuinely reproducible as tools, data, and models change?
• How can ambiguous technical failures be turned into evidence that other researchers and maintainers can inspect and act on?
• What new questions become possible when research, software, and intelligent systems are designed together?

Together, these projects exercise research design, data validation, Python and web tooling, embedded systems, packaging, CI, release engineering, and public documentation.

• Research framing — turning a broad question into a testable protocol, explicit criteria, and a result that can be challenged.
• End-to-end building — moving from reproduction and implementation through tests, packaging, CI, release, documentation, and a usable entry point.
• Work across unfamiliar systems — learning an existing codebase, locating the actual failure boundary, and making a scoped change that fits its rules.
• Evidence-aware communication — separating a proposal from an accepted result, a passing test from a general claim, and public evidence from owner-controlled metrics.

I also contribute to projects outside my own repositories. The card below counts only changes that independent upstream repositories report as merged.

The current public record contains 32 merged upstream pull requests across 18 repositories and 17 upstream owners. Representative examples include:

• TheELNFileFormat #157 — a reusable web .eln checker backed by the project's test suite.
• CycloneDX Python #1028 — encoded-path handling for XML schema loading, with regression coverage.
• Astropy #20256 — degraded-accuracy handling for expired IERS predictive values, including tests and documentation.
• Plotly.js #7959 — numeric color sorting in the parcats bundle, with implementation and tests.
• rclone #9823 — retryable handling when a successful S3 UploadPart response omits an ETag.
• Apache Magpie #1118 — more stable per-PR progress for PR-management triage and its evaluation fixtures.

I welcome conversations about research collaborations, internships, and technically ambitious engineering projects, especially where rigorous investigation and practical building belong together.

I am based in Hong Kong. In-person work in Hong Kong or Shenzhen is practical. For teams in North America, Europe, and other regions, remote collaboration is generally the most workable arrangement; long-term relocation outside Hong Kong may be difficult at present.

Email me · Explore all repositories · Published Python packages