Ship daily. Learn constantly.
We move in small increments and check them against the rubric. The work is the feedback. Big-bang releases hide the lesson; the daily push surfaces it.
We build the instrument that records what models do — eval, review, evidence, release. We are looking for engineers, designers, and specialists who would rather ship proof than slogans.
A short list, hand-picked. Each role is one we need this quarter, not a wishlist.
Engineering, design, evaluation, and field. Small enough to know every voice.
Anywhere you do your best work. Quarterly on-sites. Real equity in every offer.
We are not hiring for headcount. We are hiring people who would rather build the record than the pitch. Three things stay non-negotiable.
You will ship code that runs inside frontier labs and regulated programs. Every release leaves a record a customer can hold up in a review.
Engineers, evaluators, and reviewers who have shipped models, owned releases, and signed the proof. One platform, one record, one voice per door.
Remote-first with quarterly on-sites. Cash plus real ownership. Clear scope, fast decisions, four-step hiring loop, transparent compensation.
We do not screen for puzzle skill. We do not test trivia. The interview loop mirrors the day job: a real problem from the codebase, a real rubric being calibrated, a real release deciding whether to ship. If the work feels like the work, the conversation lands. If it does not, we are wasting your day and ours.
We move fast on decisions and slow on the questions that matter. Forty-eight hours from final round to offer. Clear feedback either way. Compensation is on the role card before we ever speak. Equity is real, vested, and modeled at the offer stage. No tricks. No tiers invented to keep the offer down. The bar is high because the work asks for it — and because the team we are joining you to is exactly that.
Final round to written offer.
No leetcode. Problems from the repo.
Salary band shown before the call.
Strike, vesting, dilution at offer.
Not values poster copy. The principles below get used on Tuesdays — in stand-ups, in design reviews, and in the moment a release wants to slip the rubric.
We move in small increments and check them against the rubric. The work is the feedback. Big-bang releases hide the lesson; the daily push surfaces it.
Every person owns the line they ship and the decision behind it. Once it is on the record, the rest of the team sees the working — the rubric, the trade-off, the audit trail.
Reading first, then a tight loop. Half the bugs come from skipping the read. The other half come from spending three weeks on a read that should have been a Tuesday.
Code that lasts is code you can read in a year. We invest in the shape of the system — the boundaries, the names, the tests — so the next change does not require a re-write.
We aim for clarity in the offer letter. Cash on the table, benefits spelled out, equity modeled. The six things below are what every full- time hire signs into on day one.
Workstation, monitor, headphones, license stack — no questions asked. If the work needs it, the company buys it.
Medical, dental, and vision at 100% for you and your dependents. Mental health and family planning included.
Untracked PTO with a four-week floor everyone clears. Two company-wide weeks closed each year so nobody pays for rest in catch-up email.
Books, courses, conferences, a paper club, a sabbatical for the team member writing the paper. No approval theatre.
Early-employee strike prices, ten-year exercise window, 409A refresh annually. We hand you the model with the offer.
Remote-first across US, EU, Canada, UK. Quarterly on-sites the team helps plan. No surveillance, no clock-watching.
Company roles build AuraOne from the inside. The specialist network routes remote experts into program-defined tracks through one shared intake. Different work, different cadence, same standard.
A small, senior team. Engineering, design, evaluation, field. Direct ownership, salary plus equity, four-step hiring loop, and a forty-eight-hour decision window.
Frontier labs run on this. You are the specialist reviewing the cases the model misses, teaching robots, calibrating rubrics, and grading the cases that count. Paid per accepted task.
Each card: the team, the work, the signal we are looking for. If you do not see your role and the work fits, write to us anyway — we read every note.
Calm systems beat heroic cleanup.
Own the evaluation spine — pipelines, routing, and audit trails that keep frontier releases predictable. TypeScript, Postgres, distributed systems.
Evidence in, signed packet out.
Build the tooling that trains, evaluates, and deploys models inside the federated lab. Regression Bank spine. Python, PyTorch, GPU infra.
Interfaces a reviewer trusts at 2 a.m.
Own the reviewer interface and annotation workbench — the surfaces specialists spend their day in. React, Next.js, TypeScript, design-system depth.
Readout over chrome. Always.
Design the evaluation experience, audit-trail readouts, and specialist onboarding. Figma fluency, B2B SaaS depth, design-system ownership.
Reviewers in context, cases in hand.
Run the reviewed record for regulated programs. Calibrate rubrics, route hard cases, and keep the signed packet honest. Domain depth, calm under pressure.
The whole loop runs in about two weeks. We tell you where you stand after every step. If we say no, you get the reason. If we say yes, the offer is on the table within forty-eight hours.
A founder or hiring manager. We share what we are working on, you share what you are working on. No prep. We listen for alignment, not buzzwords.
→Real problems from the codebase. No tricks, no leetcode. You work through the kind of question that lands in the standup the day you start.
→Meet the team. Deep dive on systems. Pair on a real change. Walk through your past work in detail. We answer every question you bring.
→Two short reference calls — you choose. While we run those, we model your equity offer and tighten the band on the role card.
→Salary, equity strike, vesting, ten-year exercise window, signing terms — all on one page. Forty-eight hours from final round to written offer.
No tricks. No leetcode. We evaluate the work you would actually do here. If a step does not produce signal we both trust, we stop and name it.
“Every interview was a question I would have wanted to argue about on a Tuesday. By the second round I knew the work. By the fifth I knew the team. The offer landed three days early.”
The things candidates actually want to know, written the way we would answer them on a phone call. If something below is not clear, write to us and we will rewrite it.
Yes, for senior roles in the US, UK, and EU. We have done H-1B transfers, O-1 conversions, and UK Skilled Worker sponsorship. The team is small but the immigration counsel is excellent.
Roughly thirty people across nine time zones. Engineering and design skew US Eastern and EU. Field and evaluation skew global. Pacific time zones are well represented for engineering.
Early-employee strike prices. Four-year vesting with a one-year cliff. Ten-year post-termination exercise window so you do not have to liquidate the cap table to keep your options. We model the offer with you.
Two short stand-ups, one design review, one team retro. The rest is heads-down work. Meetings are short and have a rubric. The default mode is async.
Greenfield with a real footprint. The platform is two years in and growing. Every hire is on a new surface, not a maintenance shift. We rotate on-call so the burden does not sit on one person.
Write anyway. If the work fits and the team needs the shape, we open the role. The five doors above are this quarter's specific needs, not a wall.
A short note. A repo, a paper, a release you owned, a review you ran. Tell us the system you would build if the day job got out of the way. We will write back within forty-eight hours.