Mistral Bought a 6-Person Language Team. Source the Rest From Git Logs
Mistral's new Language team is hiring type-system and compiler engineers. Title search misses most of them, so here is how to source them from commit history.
If you're hiring type-checker and compiler engineers, most of the people you want don't look like type-checker authors on paper. Their profile says "Software Engineer at Meta," and their real résumé is in a git log.
That search just got more competitive. On November 5, 2026, Mistral AI announced it had acquired SkipLabs, the company founded by Hack co-creator Julien Verlaguet, and brought six of its people into a new Language team under chief scientist and co-founder Guillaume Lample. Verlaguet's announcement says hiring is already open.
What did Mistral actually buy with SkipLabs?
Mistral bought a team, not a product: six people from SkipLabs, with no disclosed price and, according to Pomegra's coverage, no product with meaningful revenue and no timeline. It was Mistral's fifth acquisition of 2026.
Verlaguet's announcement names the SkipLabs team as Mehdi Bouaziz, Benno Stein, Cristiano Calcagno, Daniel Lopes and Hugo Venturini. According to his latest profile update, Verlaguet leads the Mistral AI Language team from Barcelona. SkipLabs itself was based in Paris and London. (The press doesn't agree on its age: Usine Digitale says it was founded in June 2022, Maddyness says six years ago.)
The mission is in one line from Verlaguet: today's tools were built for humans, while models need "tools fast enough to run inside their loop, feedback they can learn from." A static analyzer is a program that reasons about code without running it and catches bugs before execution. Inside an agent loop, the model writes code, the analyzer returns errors, and the model fixes them. That only works if the analyzer answers fast.
His hiring call lists five areas:
- Programming languages
- Type systems
- Compilers
- Static analysis
- Formal methods
Very few people have built a production type checker at monorepo scale, and they rarely carry a title that says so.
Why does title search miss most type-checker engineers?
Title search misses them because the company that built the most production type checkers doesn't hand out specialist titles. In Refolk's index of professional profiles, 667 US profiles carry titles matching Compiler, Static Analysis or Programming Languages Engineer. In a sample of 25 of them, Apple is the top employer with 8. Meta appears twice.
Meta has shipped three production type checkers: Hack for PHP, Flow, and Pyre (now Pyrefly) for Python. You'd expect Meta to dominate any search for type-system engineers. It doesn't, because Meta's levelling system gives nearly everyone the title "Software Engineer." Apple wins the keyword match, not the talent count.
The numbers are also noisy. The US results include a "Warehouse Operative," and the French results include profiles at KKR and Genpact. Treat every title-based count below as an upper bound.
| Segment | Count | Source |
|---|---|---|
| Compiler / Static Analysis / PL Engineer titles, US | 667 | Refolk's index |
| Same titles, France | 56 | Refolk's index |
| US to France ratio | about 11.9x | Derived |
| OCaml listed as a skill, US + UK + France | 12,301 | Refolk's index |
| Pyrefly GitHub contributors (March 2026) | 162 | Posit |
If you want to find static analysis engineers, a title filter gives you a list that leans toward Apple and leaves out most of the people who wrote Meta's checkers.
How big is Mistral's home talent pool?
It's small. France has 56 profiles with specialist PL titles, about one-twelfth of the US count, and only 4 of a 25-profile sample are in the Paris region. The sample is spread across AdaCore, Embecosm, Kalray and Mozilla, and across Grenoble, Strasbourg and Lyon.
The geography tells you how this team is likely to hire:
- Across Europe and remotely. The lead is in Barcelona, and Paris alone can't staff a language team.
- From London. SkipLabs had a London office, and London-based Meta alumni are the obvious people to poach.
For anyone looking at programming language engineer jobs from the candidate side, the same math applies. Demand just went up and the supply hasn't changed.
Where do the real candidates show up?
They show up in commit history. The best sourcing channel for type-system engineers is the contributor graph of the projects they built, not their profile headline. Six repositories cover most of the production experience that matters:
- Hack (Meta), the PHP type checker, written in OCaml.
- Flow (Meta), also written in OCaml.
- Pyre and Pyrefly (Meta). Pyre was OCaml, and Pyrefly is the Rust rewrite.
- Sorbet
- mypy
- TypeScript
Some of the work is already done for you. Pyrefly's launch post credits 15 engineers by name, including Jia Chen, Rebecca Chen, Sam Goldman, Neil Mitchell, Maggie Moss, Steven Troxler, Lucian Wischik, Danny Yang and Sam Zhou. That's a ready-made shortlist that no title search would return.
There's some irony in it, too. When Hack's team was debugging the type checker across Facebook's PHP codebase, they replayed Facebook's git history. The tool was debugged with git history, and you can find its builders the same way.
What to read in a commit log
Commit counts alone aren't enough. Look for these signals:
- Commits to the checker core, meaning inference, the solver, the incremental engine. Docs and test fixtures count for less.
- How long they contributed. Two years of steady commits says more than one big month.
- Review authority. People who approve other people's pull requests usually own a subsystem.
- Which files they touched. Shared memory, caching and parallelism code points to the speed skills Mistral wants.
- When they stopped. A contributor whose commits stop in a given month has probably moved teams or companies.
Doing this by hand across six repos is slow. The gap between "who wrote the solver in Flow" and "what is that person's current profile and email" is the slow part, and it's the gap Refolk closes: you describe the person in plain English and get back a ranked shortlist that joins GitHub activity to LinkedIn and the open web.
Is OCaml a better signal than a job title?
Yes, but it pulls in a different crowd. In Refolk's index, 12,301 profiles across the US, UK and France list OCaml as a skill, roughly 17 times the number of specialist-titled profiles (a directional figure only, since the OCaml set includes the UK). The top titles in that group aren't "Compiler Engineer." They're Member of Technical Staff, CTO, Co-Founder and Founder.
The reason it works: Hack, Pyre and Flow were all written in OCaml, so anyone who spent years inside those codebases had to learn it. The OCaml group already includes people with real type-checker experience, many of them under generic titles. Anthropic and Sec3 both show up among the employers, and Member of Technical Staff is the title frontier labs use, so some of these people already work at Mistral's competitors.
The catch is the founders and CTOs. For a lab, these are often people you're competing against, not people you can hire. A founder is more likely to be an acquisition target, which is exactly how Mistral got Verlaguet.
The fix is to stack signals. Use OCaml as a skill, then require at least one commit to a checker repository. That cuts out the generalists and keeps the people who actually shipped a checker.
What does the Rust rewrite tell a sourcer?
The move from OCaml to Rust widened the contributor pool, and the git log shows exactly when each new person joined. Anyone who started contributing to Pyrefly after the rewrite is a newer and mostly untapped candidate.
Neil Mitchell, who worked on Buck2 before Pyrefly, put the OCaml problem bluntly in a Jane Street talk: Pyre would get only "one or two people who took enough time to learn OCaml" as outside contributors. Rust removed that barrier. By March 2026, Pyrefly had 162 GitHub contributors. Speed improved too: Meta reports that Pyrefly checks the Instagram codebase in 13.4 seconds, compared with more than 100 seconds for Pyre.
The Python type-checker field gives you four contributor pools of very different sizes:
| Python type checker | GitHub contributors (March 2026) |
|---|---|
| ty | 186 |
| Pyrefly | 162 |
| Basedpyright | 82 |
| Zuban | 17 |
Small pools are easier to work through. With Zuban's 17 contributors, you can read every one of them in an afternoon. With Pyrefly's 162, sort by first-commit date and look at the people who joined after the rewrite and stayed. They've proven they can work in a production checker, and few recruiters have reached them yet.
Why does incremental computation matter more than type theory?
Mistral needs checkers fast enough to run inside a model's loop, so the scarcest skill is incremental computation, not type theory. Incremental computation means recomputing only what changed after an edit instead of re-analyzing the whole codebase.
Hack shows why this matters. Starting around 2011, it added a static type system to a PHP codebase of about 30 million lines, using a multi-process, shared-memory OCaml architecture. That is the same engineering problem an agent loop creates, just with a model making the edits.
So widen the profile beyond people who publish on type soundness. Look for people who have worked on:
- Salsa-style incremental query engines.
- Buck2, where Mitchell worked before Pyrefly.
- Hack's shared heap.
This is why these searches drag. The people with both type-system depth and incremental-engine experience mostly carry generic titles, and every lab building coding agents can read the same Pyrefly credits page you can. Working the same short list as everyone else is how a single role stays open far longer than planned.
A 30-day plan for recruiting type-systems engineers
Start with commit history, verify with profiles, and save title search for last. Here's the order I'd follow for a type-systems search starting this month:
- Week 1: Collect names from credits. Pull the 15 names from the Pyrefly launch post and the named SkipLabs team, so you know who's already taken.
- Week 1: Map the six repos. Export contributors for Hack, Flow, Pyrefly, Sorbet, mypy and TypeScript, filtered to core checker paths.
- Week 2: Sort by tenure and recency. Separate long-time core contributors from people who joined Pyrefly after the Rust rewrite.
- Week 2: Cross-reference with OCaml and Rust skills. Keep anyone who lists OCaml or Rust and has commits to a checker. Drop pure founders unless you're scouting for acquisitions.
- Week 3: Check location. If you're competing with Mistral, focus on London and the French regional hubs (Grenoble, Strasbourg, Lyon) before Paris.
- Week 4: Write outreach that names the commit. "You rewrote Flow's incremental recheck" gets a reply. "Exciting AI opportunity" doesn't.
Steps 2 through 5 are where most of the time goes. With Refolk, you can ask for the whole stack in one sentence, for example core contributors to a named checker, with Rust or OCaml, living in Europe, and get the joined profiles back instead of building the spreadsheet yourself.
FAQ
How do I hire compiler engineers whose profiles just say "Software Engineer"?
Search the code, not the title. Meta built Hack, Flow and Pyrefly, yet only 2 of 25 sampled US specialist-titled profiles in Refolk's index work there, because Meta titles nearly everyone "Software Engineer." Start from the contributor graphs of checker repositories and from published credits like Pyrefly's 15-name list, then match those GitHub identities to current profiles.
Is the Mistral AI Language team hiring outside France?
Almost certainly in practice. Verlaguet leads the team from Barcelona, SkipLabs had offices in Paris and London, and France has only 56 profiles with specialist PL titles in Refolk's index. A team this specialized will have to hire across Europe and recruit remotely.
Is OCaml experience required for type-system roles?
No, but it's a strong signal. Hack, Flow and Pyre were written in OCaml, so anyone who spent years in those codebases had to learn it. Pyrefly is written in Rust, and that rewrite brought in far more contributors (162 by March 2026). Ask for experience with a production checker, not a specific language.
How many type-system and static analysis engineers are there?
Fewer than any keyword search suggests. Refolk's index shows 667 US and 56 French profiles with matching titles, and both figures include noise, so treat them as upper bounds. The real pool is the people who have committed to the core of a production checker, and you can count those directly in each repository's contributor graph.
The data behind this
Try it on the search you came here for
Stop building boolean strings. Just describe the person.
Type one sentence. I plan the search, read GitHub, public LinkedIn and Crunchbase records, and the open web as it is right now, and hand back a ranked list with the reason next to every name.
01Describe them
One plain sentence. Role, city, stack, stage, whatever matters to you.
02I read the web live
GitHub, public LinkedIn and Crunchbase records, the open web. Not a database that went stale last quarter.
03You read the shortlist
Ranked, with the reasoning under every name. Open a profile, ask a follow-up, narrow it down.
- Staff backend engineers in NYC who shipped Rust in production
- Series A fintechs in SF under 50 people, growing headcount this year
- Maintainers of fast-growing Rust web frameworks on GitHub
- No boolean, no filters, no seat to buy. One box.
- Read at search time, so a profile updated yesterday counts today.
- Every step visible as it runs, every name with its reason.
500 free credits on sign-up. No card, no demo call. See real searches.