Training data from professional engineers.

Professionals

Engineers with at least three years of production experience, and six for code review.

Vetted

Each one passes a recorded technical interview and a paid trial task before project work.

Matched

We assign engineers by stack and domain to fit your specification.

Paid properly

Published hourly rates attract experienced people and keep them on your project.

Bug fixes

A repository with a reproducible bug, the issue, failing tests and a verified fix.

Features

An existing repository, requirements, acceptance tests and a working implementation.

Code review

Proposed changes, the defects an experienced reviewer would flag, and tested corrections.

Migrations

Original code, the migration requirement and a verified updated version.

Trajectories

Recorded engineering sessions: inspecting files, running commands, editing and testing.

We currently focus on TypeScript and Node.js backends.

Private link

Each batch arrives as a direct, private download link to the data and its manifest. Links expire, and we issue a new one whenever you need it.

Your storage

Or we deliver straight into your own cloud storage bucket.

Format

Repositories as archives or container images, records as JSONL. We can match your schema.

Manifest

Every batch lists each task with its author, date, licence and file checksums.

Reproducible

Each task runs from a clean setup and the bug reproduces reliably.

Verified

Reference solutions pass the failing tests and all regression tests.

Original

Tasks are written for the engagement, not taken from public repositories or benchmarks.

Each record lists its author, date and licence. Exclusive work is priced separately from work we may license to others, and evaluation sets are kept apart from training data. Data handling.

1. 1
   SpecificationFormat, stack, difficulty, permitted sources and acceptance criteria.
2. 2
   PilotTen tasks, paid upfront, to agree the quality bar.
3. 3
   DeliveryOngoing batches, each with a manifest.