Selected work · Open source · Achievements · 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 across disciplines. My work brings together research, engineering, experimentation, and open collaboration. I am currently exploring frontier AI: how advanced systems reason and fail, how we should evaluate them, and how they can become useful in scientific and engineering work.
I like turning promising ideas into things other people can inspect: runnable software, explicit tests, reproducible evidence, public releases, and clear records of results and open questions.
- The Chinese University of Hong Kong — PhD researcher in Electronic Engineering.
- Nanjing University — bachelor's degree.
A few questions currently keep me thinking:
- 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?
These are the projects I want people to inspect. Several earlier experiments are archived rather than presented as products.
| Project | What you can inspect | Public entry points |
|---|---|---|
| STM32 multifunction robot car | Physical embedded system: line tracking, obstacle avoidance, ultrasonic ranging, Bluetooth control, servo scan, OLED. | source · v0.1.1 |
| BenchLineage | Experiment provenance, instrument identity, calibration, uncertainty budgets, evidence bundles, and ELN exchange. | PyPI · v0.3.8 |
| FrontierTrials | Private blind comparisons of AI products on your own work. No API keys. | try it · study report |
| OhmJudge | Fresh EE calculation tasks with executable local grading. Early research tool, not a model ranking. | collector · methodology |
| ReproWeave | Evidence maps and replication triage for papers, with clearly marked synthetic demo data. | project · v0.4.2 |
| TermScope | Terminal serial plotter for Arduino, ESP32, and STM32 over UART, pipes, or SSH. | PyPI · readme |
Other maintained tools
- ColdShelf — private catalogue for unplugged drives.
- VulnFuse — explainable correlation across SARIF, SBOM, and scanner findings.
- contextcost — measure repository token cost and verify that cuts save tokens.
- EvalInt — integrity checks for reference-scored LLM eval sets.
- WillItBreak — breaking API changes that actually reach your call sites.
- 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 — making assumptions, results, and unresolved questions visible so collaborators can inspect and challenge them.
I contribute to foundational tools, scientific computing, and security frameworks. As of 2026-09-10, the public record contains 37 merged upstream pull requests across 22 repositories and 21 upstream owners. The full receipt list is in CONTRIBUTIONS.md.
Tip
Live Upstream Verification: All external contributions are public, inspectable, and verifiable directly on GitHub. 🔗 Explore All Merged Upstream PRs via GitHub Search →
| Target Ecosystem / Repository | Upstream PR & Receipt | Technical Scope & Impact | Status |
|---|---|---|---|
Apache Foundation (apache/magpie) |
#1118 · e8c3a346 |
Per-PR progress headers for triage loop & 4 test fixtures | Merged |
GitHub Official (github-mcp-server) |
#3146 · ffa16d00 |
Feature flags via URL query parameter for headerless connections | Merged |
PyPA (pypa/cibuildwheel) |
#2977 · 3d9e8c5c |
Refreshed CPython & CircleCI platform documentation cross-links | Merged |
Plotly (plotly/plotly.js) |
#7981 · #7959 | Cartesian hovermode cursor spike behavior & numeric color sorting | Merged |
Astropy Core (astropy/astropy) |
#20256 · 9f4de8d6 & #20359 |
Degraded-accuracy IERS handling & Table primary key engine | Merged / Active |
CycloneDX (cyclonedx-python-lib) |
#1028 | Encoded-path handling for XML schema loading with regression tests | Merged |
Rclone (rclone/rclone) |
#9823 · #9818 | Retryable error handling for S3 UploadPart missing ETag response |
Merged |
Restic (restic/restic) |
#22029 · #22028 | Removed contradictory plaintext exception; updated crypto docs | Merged |
The ELN Consortium (TheELNFileFormat) |
#157 | Reusable web .eln format validator backed by test suite |
Merged |
Contribution record and current activity
The same snapshot includes 69 open external proposals across 46 repositories and 41 upstream owners. The merged total above records work already accepted by upstream repositories.
The accepted set is generated from a versioned manifest and checked against live GitHub state. Detailed PR receipts, reviews, and dated updates are recorded in CONTRIBUTIONS.md and COMMUNITY_FOOTPRINT.md.
My GitHub profile activity is backed by verified engineering receipts, formal badge qualifications, and upstream community recognition:
| Achievement | Tier / Status | Verifiable Receipt & Scope |
|---|---|---|
| 🧠 Galaxy Brain | Accepted Answers | Maintainer-accepted answers resolving architectural and runtime questions in beetbox/beets #6981 and fastapi/typer #1954. |
| 🦈 Pull Shark | Gold Tier | Over 54 merged pull requests across global foundation ecosystems (Apache, OpenSSF, PyPA, Astropy, Plotly, CycloneDX). |
| ⚡ Quickdraw | Sub-Minute Triage | Automated, sub-minute issue triage and closed-loop resolution (CAOShurong#102). |
| 👥 Pair Extraordinaire | Verified | Co-authored commits across distributed upstream open-source repositories and cross-team maintenance. |
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