Purpose
Pretraining, fine tuning, model evaluation or another defined research use. Explain what the model should learn and what would make a collection useful to your team.
Research data / For AI and research teams
Source data for
your next training run.
Tell us the scientific problem, model task and collection your team needs. Include a target scale if you are planning a large pretraining corpus, a focused fine tuning set or an independent evaluation collection.
A useful sourcing brief
Describe the science.
Then define the scope.
A short initial description is enough to begin. The following points help develop a requirement that potential data holders can assess.
Scientific coverage
Materials or life science area, inclusion criteria and relevant experimental range. Identify which parts of the requirement are essential and which could be flexible.
Measurements and context
Raw instrument files, images, structured measurements and any paired modalities. Describe the sample identities, methods, labels, controls and outcomes you need.
Scale
Approximate bytes, independent samples or experiments and minimum useful scope. If the target is terabyte scale, describe the scientific breadth behind that volume.
Evaluation
The formats and record relationships your team needs to inspect, the criteria for acceptance and the appropriate size of an initial evaluation sample.
Use and timing
Intended research or commercial training use, evaluation milestones and delivery needs. Permitted use and transfer arrangements are addressed for the specific collection.
Two example briefs
Make the requirement concrete.
Illustrative requirement
Materials and process data
Preparation → characterization → properties
We are developing a model that connects material preparation with measured properties. We need linked process records, raw characterization and property results, including unsuccessful experiments where available. We would like to evaluate a small linked sample before discussing a collection measured in terabytes.
Illustrative requirement
Life science research data
Molecular measurements + research imaging
We are developing a model that combines molecular measurements with research imaging. We need linked sample identifiers, preparation methods, annotations and experimental outcomes. We would like to understand study coverage and review an agreed sample before defining the training collection.
The acquisition process
From requirement
to an agreed collection.
SIB coordinates sourcing and the licensing discussion around your brief. Availability and suitability are established through qualification.
Define the brief
Clarify the scientific objective, data types, essential context, target scope and intended use.
Qualify potential sources
Identify relevant data holders and discuss existing records, their coverage and open questions.
Evaluate the fit
Agree the description or sample needed for your technical review, including the recipient, access and handling arrangements.
Agree the transaction
Define the collection, permitted use, acceptance criteria, preparation work and delivery before a transfer.
What does your model
need to learn?
Send a brief description of the requirement. There is no need to complete every technical detail before the first conversation.