For AI companies
Data for the next generation of AI.
We source, assess and prepare licensed datasets around specific model development and evaluation requirements, from rights holders with the authority to license them.
- application
- Coding-agent evaluation
- source
- Software company archive, licensed
- context
- Issue, review thread, merged change
- volume
- ~20,000 resolved issues
- languages
- Python, TypeScript
- format
- JSONL, one issue thread per line
- identifiers
- Names, emails and secrets removed
- licence
- Evaluation only, no redistribution
What we source
Data we source to your brief. Every dataset is assessed for rights and suitability before it is offered for licence.
Publishing and text archives
Books, journals, trade press and specialist archives.
- Full text with its structure: chapters, sections, references
- Bibliographic and subject metadata
- Editorial and revision history
Video and footage archives
Unused rushes, documentary and educational archives, and specialist footage.
- Original resolution and frame rate
- Shot and scene metadata
- Linked audio tracks and transcripts
Software and engineering records
Codebases, issue histories, code review and incident records.
- Links between issue, discussion and change
- Review comments in sequence
- Test suites and build history
Workplace records
Communications, tickets and workflow histories from the tools teams use.
- Thread structure and reply order
- Handoffs and approvals between roles
- Linked tickets, documents and decisions
Customer support records
Tickets, chat transcripts and help-desk histories.
- Issue category and priority
- Resolution path and escalations
- Agent and customer turns
Documentation and knowledge bases
Manuals, wikis, policies, process documentation and FAQs.
- Document structure and hierarchy
- Revision history
- Cross-references
Expert and professional input
Answers, demonstrations and evaluations written by qualified professionals.
- Credentials and specialism of each contributor
- Reasoning alongside final answers
- Rubrics and reviewer agreement
Speech and conversational audio
Consented recordings of natural conversation across UK accents and dialects.
- Speaker metadata such as accent and region
- Verified transcripts and timestamps
- Speaker turns and overlaps
Built around your requirements
We start from your brief, not a catalogue. The more precisely you describe what the data has to do, the better we can judge whether a partner's data fits, and the less preparation is wasted.
Application
Training, fine-tuning, evaluation, agent development or research.
Content
The material the data must contain: documents, footage, code, conversations or workflows.
Context
Structure that must survive preparation, such as metadata, threads, links or outcomes.
Scale
Approximate volume, time span and any language or domain mix.
Format
Preferred schema and file format, for example JSONL with one record per line.
Quality
Thresholds for completeness, noise, duplication or labelling.
Licensing
Intended use, exclusivity, term and any redistribution needs.
Provenance and permitted use
Every dataset is assessed for provenance, rights, restrictions and suitability before it is offered for licence. If those can't be established, the dataset isn't offered.
For each licensed dataset, we aim to provide documentation covering:
Source description: the type of organisation and systems the data came from
Collection period and coverage
A summary of the authorisation basis and the uses the data partner has agreed to
Processing applied, including filtering and de-identification methods
Residual risks and known limitations
Permitted uses, restrictions and obligations under the licence
Licences are purpose-specific. A dataset licensed for evaluation can't simply be reused for training. Tell us your full intended use up front so the terms cover it.
From requirement to delivery
Timelines depend on the data and the rights assessment involved. We give you a clear view of feasibility early, before either side invests heavily.
Requirements
You describe the application, interaction types, scale, format and licensing needs.
Sourcing
We look for data partners whose data could plausibly meet the brief.
Assessment
Rights, permitted uses, privacy and suitability are reviewed before anything is prepared.
Human review
Preparation
The agreed scope is extracted, structured, de-identified where appropriate and checked.
Contracting
Licence terms set out the dataset, permitted uses, restrictions and obligations.
Delivery
The dataset and its documentation are delivered through an agreed secure method.
Discuss your requirements
Share what you need at whatever level of detail you have. We'll come back with questions and an initial view on feasibility.
- Your requirements are treated as confidential
- We source to your brief, so specifics help
- Submitting doesn't commit you to anything