Corpix Labs · Data for AI Infra

Spec it.We source it.

Corpix sources the training data AI labs can’t get anywhere else, straight from the companies and experts who hold it. Tell us what you need, or see what your own data is worth to AI labs.

Tell us your requirement →Scan your data →

The data your model needs is locked inside companies. We unlock it.

On the shelf

Five lines, ready now.

01 / 05
CODOn the shelf01 / 05

Code

Real engineering, not tutorial code.

  • Full commit history and diffs
  • Code review threads and issue trackers
  • Incident write-ups and runbooks
Ships as
Git bundles · JSONL
Starts with
A test sample your team runs before signing
Request a sample → Details
OPSOn the shelf02 / 05

Company operations

How work actually gets done.

  • Tickets, workflows and approvals
  • Internal docs and SOPs
  • Tool-use and decision logs
Ships as
JSONL · Parquet
Starts with
A test sample your team runs before signing
Request a sample → Details
EGOOn the shelf03 / 05

Egocentric video

Hands-on footage from real shop floors.

  • First-person capture across industries
  • Task and action annotations
  • Our capture rig or yours
Ships as
MP4 + annotations · WebDataset
Starts with
A test sample your team runs before signing
Request a sample → Details
MEDOn the shelf04 / 05

Medical

Patient journeys, measured over years.

  • Longitudinal records across specialties
  • Labs, notes and treatment timelines
  • De-identified before transfer
Ships as
Parquet · JSONL
Starts with
A test sample your team runs before signing
Request a sample → Details
AUDOn the shelf05 / 05

Audio

Speech in the languages models miss.

  • Indic and low-resource languages
  • Transcripts, alignment and speaker labels
  • Conversational and read speech
Ships as
WAV / FLAC + transcripts
Starts with
A test sample your team runs before signing
Request a sample → Details

The shelf, at a glance

Pick one. Try a sample.

Full catalogue →
01 Code Real engineering, not tutorial code. Git bundles · JSONLExplore → 02 Company operations How work actually gets done. JSONL · ParquetExplore → 03 Egocentric video Hands-on footage from real shop floors. MP4 + annotations · WebDatasetExplore → 04 Medical Patient journeys, measured over years. Parquet · JSONLExplore → 05 Audio Speech in the languages models miss. WAV / FLAC + transcriptsExplore → +4 Sourced on demand Not on the shelf? We’ll source it for you. CAD · decks · models · feedbackSee them ↓

Sourced on demand

Not on the shelf? We’ll source it.

We go to the firms and experts who hold it, and source it to your schema, volume and quality bar.

  • Parametric parts and assemblies
  • Feature trees and drawings
  • From real engineering teams
STEP · native filesRequest sourcing →
  • Consulting, banking, board and sales decks
  • Editable native files
  • From working professional teams
PPTX · PDF + JSONRequest sourcing →
  • 3-statement, DCF and LBO models
  • Live formulas and linked tabs
  • From working finance teams
XLSXRequest sourcing →
  • Preference pairs and rankings
  • Rubric grading and written critiques
  • Multi-turn expert trajectories
JSONLRequest sourcing →

What’s your model missing? Tap it in.

Finish your requirement → Or talk it through

What does your org hold? Tap it in.

Finish the 3-minute scan → How licensing works
Your requirementtap a line
Available now—
  • Your selection appears here.

How an order runs

Brief. Sample. Sign. Ship.

The full process
  1. 01

    Brief

    30 minutes on what the model needs.

  2. 02

    Sample

    Real data from the set, tested on your workloads.

  3. 03

    Sign

    Plain terms built around your use.

  4. 04

    Ship

    Your formats, documented file by file.

Need data, or
have some?

Two minutes on your requirement, three on a data scan, or 30 on a call. Every route ends with next steps.

Tell us your requirement →Scan your data →