HATCH by NOBLESTARK
Get paid to work on the data behind serious AI.
Hatch is NobleStark's expert network of 5,000+ vetted specialists across software engineering, cybersecurity, math, code, reasoning, and more. Get paid top-of-market to build the data, benchmarks, and evaluations that train and measure AI systems — from enterprise ML to the world's leading labs.
- Top-of-market rates, weekly payouts, every engagement under a clean contractor agreement with IP and NDA scoped per project.
- Work on the data that trains frontier models — not synthetic tasks, not busywork.
- Fully remote, async-friendly, and you choose your hours. Pick up shifts on open tracks or join a long-running embedded pod.
Why experts choose Hatch
Hatch exists because the best data is produced by the best people. We pay them accordingly, contract them cleanly, and give them work that actually moves the frontier — not annotation farm rates dressed up in nicer marketing.
Top-of-market pay
Senior engineers earn $80–$200/hr depending on track and seniority. Math PhDs and cybersecurity specialists in scarce verticals earn more. Rates are public on each track; no opaque markups, no per-task races to the bottom.
High-impact projects
Real evals, real benchmarks, real preference data shipped to companies building serious AI — from enterprise ML teams to the world's leading research labs. Not synthetic busywork; work that actually ships.
Remote, async, your hours
Pick up shifts on open tracks or join an embedded pod with a steady weekly commitment. We accommodate timezones from PT to IST. Async-friendly review pipelines so you can do deep work uninterrupted.
Weekly payouts
Approved work is invoiced and paid weekly. Stripe, ACH, wire, or Wise — whichever works in your country. No 60-day procurement cycles.
Clean contracts. Your name on the work.
Every engagement runs under a contractor agreement with scoped IP and a project-specific NDA. Every item you produce is attributed to you in our delivery provenance — your work is visible to the lab consuming it.
Designed by practitioners
Hatch's review pipelines, calibration tasks, and rubrics were built by people who've shipped the same work — senior engineers, eval leads, and domain specialists. That's the standard we hold every applicant to.
Where Hatch experts work
Five tracks cover the verticals that move frontier model performance the most. Apply once — we'll match you to any track you're qualified for, and you can take on as much or as little work as you want.
Software Engineering Agentic
5+ years shipping production software. Comfortable in at least two of: Python, TypeScript, Go, Rust, Java, C++. Reads diffs faster than most people read prose.
- SWE-bench-style agentic coding tasks against real repos
- Repo-level refactors and multi-file edits
- Failing-test reproduction, dependency-aware patches
- Tool-use traces (shell, editor, debugger) with step labels
Code Generation & Review
Strong engineer with a sharp sense for idiomatic code in your language of choice. Senior code reviewers welcome.
- Multi-language code with executable test oracles
- Difficulty-banded problem authoring (junior → staff)
- Style-graded reference solutions
- Code review and static-analysis labeling
Cybersecurity & Red-Team
OSCP / PNPT / CRTO, AppSec engineer, CTF top-100 player, or equivalent industry pen-test background. Defensive specialists also welcome.
- Adversarial prompt authoring against modern model policies
- Vulnerability writeups with CWE tags and executable PoCs
- CTF transcripts: web, pwn, crypto, reverse, forensics
- SOC triage data and IOC clustering
Math & Formal Reasoning
Olympiad medalist, PhD or PhD candidate in mathematics, or theorem-prover engineer (Lean 4 / Coq / Isabelle).
- IMO / Putnam / IOI-style problem authoring with verified solutions
- Lean 4 / Coq formal proofs
- Step-level annotation of chain-of-thought reasoning
- Symbolic and numerical verification harnesses
Custom Eval & Benchmark Authoring
Eval engineer or research scientist with experience shipping benchmarks. You'll co-design and lead authoring teams for a customer's bespoke eval.
- Co-design with the customer's research team
- Item authoring playbooks, graders, leaderboard hosting
- Held-out integrity protocols, contamination prevention
- Manage 5–25 expert authors as a track lead
Don't see your domain?
We open new tracks every quarter based on customer demand. If you're senior in a domain that touches AI capability or safety, apply anyway and tell us about it.
Built by experts, paid like it
Names abbreviated at experts' request. Full references available on hire.
How vetting works
Four steps, roughly one week end-to-end. Every step is paid from the trial task onward. We tell you up front if a track isn't a fit — we don't waste your time stringing you along.
1. Apply
90-second form. Tell us your domain, credentials, and how to reach you. We respond within 48 hours.
2. Skill task
A short, paid trial task in your domain — typically 2–3 hours. Designed so we can see your real work, not just your resume.
3. Live review
30-minute review with a NobleStark domain lead. We discuss your trial submission, your strengths, and the tracks you'd be a fit for.
4. Onboard & match
Sign your contractor agreement and project NDA. Your first paid match arrives within a week. Weekly payouts start with your first delivered batch.
Our promise to every Hatch expert
What we do
- • Pay top of market and pay weekly
- • Attribute every item you produce to you in delivery provenance
- • Sign clean contractor agreements with scoped IP and NDA
- • Match you only to tracks you're qualified for
- • Give honest feedback when something is or isn't a fit
- • Protect your privacy — see our Expert Privacy Notice
What we don't do
- • Pay annotation-farm rates dressed up in nicer marketing
- • Pit experts against each other in piecework races
- • Train internal models on your work without explicit consent
- • Share your identity with customers unless you opt in
- • Drag out payments or hide behind procurement cycles
- • Match you to work that doesn't actually need your level of expertise
Questions? Email us at hatch@noblestark.com · Expert Privacy Notice