Job Description
Mercor is seeking current or former CTOs or Founders with direct access and authority over their organization’s private codebase to partner on high‑impact AI training and evaluation initiatives. You’ll be working with us and the AI company we engage to securely use your code to build realistic evals and, where permitted, train / improve coding models.
What you’ll do
- Authorize & enable access: Verify you have the legal authority to grant a limited license for evaluation and (if agreed) training use; coordinate secure repository access (read‑only, scoped).
- Curate high‑value datasets: Select representative modules / services, strip or mask secrets / PII, and define inclusion / exclusion rules.
- Design realistic evaluations: Create task suites spanning your stack (e.g., refactors, bug‑fixes, feature diffs, tests, performance work), grounded in your production code.
- Develop test harnesses: Build / guide test cases and scoring that reflect real engineering constraints (builds, CI, style, coverage, perf).
- Analyze outcomes: Review model behavior on your code, highlight failure modes, and propose improvements aligned to your architecture.
- Ensure compliance & security: Partner with us on data handling (DPA / MSA / SOW), audit logs, redaction, secret scanning, and access revocation.
You’re a great fit if you
- Are a CTO / founder (current or former) who controls or can authorize access to a substantial, production codebase (mono‑repo or multi‑service).
- Can grant a time‑boxed, purpose‑limited license for evaluation and—if mutually agreed—training on selected code segments.
- Have deep knowledge of your system’s architecture, tooling, and standards (Git, CI / CD, testing, observability).
- Are fluent in one or more of: Python, Java, C / C++, JavaScript, TypeScript (others welcome).
- Can collaborate asynchronously with researchers / engineers and move quickly with minimal oversight.
- Are an active OSS contributor on Git.
Role details
- Engagement: Part‑time, flexible, 100% remote / asynchronous.
- Scope: Evaluation‑only by default; training usage optional and controlled by written agreement (purpose‑limited, revocable).
- Security: Read‑only, least‑privilege access; optional on‑prem / VPC workflows; comprehensive auditability.
Compensation & legal
You’ll be engaged as an hourly contractor via Mercor; payments weekly via Stripe Connect.
For qualifying partners, we can structure separate, limited licenses (with fees) for evaluation and / or training use of selected code assets.
All work is governed by mutually executed MSA / SOW and data protection terms.