Verifying a Chinese engineer's credentials

Degrees: there is an official answer

China has a central verification system, which makes this easier than most countries rather than harder. Academic records for Chinese institutions are held by CHESICC — the China Higher Education Student Information and Career Center, the body behind chsi.com.cn — and it issues verification reports in English intended for foreign employers and universities.

The candidate requests the report and it comes to you from CHESICC rather than from them, which is the entire point: a report forwarded by the candidate as a PDF is worth very little, and one you receive through the verification channel is worth a great deal. It covers enrolment, graduation and the degree awarded.

Two practical notes. It takes days rather than minutes, so request it when you shortlist rather than when you are ready to offer. And UK ENIC can tell you what a given Chinese qualification is comparable to in UK terms, which is a different question from whether it is genuine and occasionally the one you actually needed answered.

Published work: verify the claim, not the CV

For research-weighted hires the publication record is usually the strongest evidence available, and it is public. The check is straightforward and worth doing properly:

• Find the paper independently. Search for the title rather than opening a supplied link. Confirm the venue exists and is what the CV implies.
• Check the author position. First author, joint first, middle, last — in most engineering and ML venues these mean materially different levels of contribution, and "co-authored a paper at NeurIPS" can mean either a great deal or very little.
• Ask them to explain the part they did not do. Someone who genuinely worked on a paper can tell you clearly which parts were theirs, which were a colleague's, and what they would do differently now. This single question separates real contributors from names on a list faster than anything else.
• Look at citations sceptically. Citation counts reward age and subject area, not quality. A 2024 paper with eight citations may be better work than a 2019 paper with two hundred.

Repositories: provenance is visible if you look

Public code is the other verifiable artefact, and commit history is hard to fake convincingly. Worth checking:

• Does the contribution history look like work, or like a burst of activity shortly before the job search? Real engineering leaves an uneven, long trail.
• Are the commits substantive, or formatting and dependency bumps? Open two or three of the largest and read them.
• Does the code match the claimed seniority? A candidate presenting as a principal engineer whose public work is all tutorial-shaped is worth a longer conversation.
• Are there code reviews they gave to others? Reviewing well is a better seniority signal than writing well, and it is visible on most public projects.

Note the limits. Plenty of excellent engineers at large Chinese platforms have almost no public code, because their work lives in internal monorepos under agreements that prevent sharing it. An empty GitHub profile is not evidence of weakness — it is an absence of evidence, and the technical interview has to carry more weight when that is the case.

Employment history, which is the genuinely hard part

This is where cross-border checking gets thin, and it is worth being honest rather than implying a rigour that is not available. Reference calls across a language barrier and a seven-hour time difference are often impractical, and Chinese employers are frequently reluctant to give substantive references at all — as many UK employers now are.

What tends to work instead:

• Specific, checkable claims. "I worked on recommendations at a large platform" is unverifiable. "I owned the candidate-generation service, it served this much traffic, here is the architecture and here is what broke" is a claim that a technical interview can test directly.
• Third-party background checks with China coverage. Several international providers verify Chinese employment records, and their coverage is genuinely better than what you can achieve by calling. It costs money and takes days; for a senior hire it is proportionate.
• Weight the work sample. Where history is hard to verify, what somebody can demonstrably do carries more of the decision. That is not a compromise — for engineering roles it is arguably the better signal anyway.

Interview signals that cross borders, and ones that do not

Assessing engineers in a second language has a specific failure mode: fluency gets mistaken for ability in both directions. A candidate who explains a distributed systems trade-off haltingly may understand it better than one who narrates it smoothly.

What travels well is anything with a right answer and a visible process: debugging something real, reading unfamiliar code aloud and explaining it, working through a design with constraints that change halfway. What travels badly is rapport, culture fit as usually assessed, and anything depending on idiom or pop-cultural reference.

Give written questions in advance where the content allows. It costs nothing, removes a language penalty that has no bearing on the job, and produces a better-signal conversation for everyone.

What we do before you see anyone

We check the degree through CHESICC, verify claimed publications and repository contributions independently, and put specific technical claims to the candidate before they reach your shortlist. What we do not do is decide whether they are good enough for your team — you run your own process and make your own call, and the engagement model is deliberately built that way round.

If a check comes back inconclusive we will tell you it was inconclusive rather than quietly leave it off the summary. An unverifiable claim is a thing you should know about, not a gap to paper over.