Comans Data Foundry
Proprietary datasets for AI model development.
Comans Data Foundry is a specialist dataset publisher for frontier AI labs and other model developers. We’re building a catalogue of proprietary data for model training and evaluation, with acquisition, creation, rights review and quality checks built into the publishing process.
Our first dataset, COM-ITOPS-001, is in development. We’re building towards a catalogue with new and refreshed datasets released each month.
- Create or acquireClear the rights
- Clean and structureProve quality and provenance
- Package and publishDefine the dataset and market the release
- License and refreshPlanned monthly releases
The business model we’re developing.
Dataset products for model developers
We’re building a catalogue of proprietary datasets for model training, post-training and evaluation. Each release will have a defined scope, dataset identifier, version, record types, quality evidence and licence terms.
First dataset in development
COM-ITOPS-001
Enterprise IT agent training dataset
In developmentCOM-ITOPS-001 focuses on the work IT teams handle every day, including identity and access, networking, Microsoft 365, monitoring, backup and recovery.
The planned dataset combines tasks, expert actions and observations, and outcome checks for agent training and evaluation. Coverage, record counts, formats, quality evidence and licensing terms will be defined for each release.
Explore COM-ITOPS-001How buyers will access the catalogue
Choose a dataset
Choose a defined dataset from the catalogue, with its scope, record types, quality evidence and licence terms set out before purchase.
License regular releases
License a series of new or refreshed datasets. We’re planning a monthly publishing cycle, with coverage, availability and commercial terms confirmed for each release.
Request a data category
Tell us which category of training or evaluation data you need. Your requirements can help shape future releases.
Potential future datasets
Working categories and identifiers for future releases. These are planning examples; scope and release dates are still to be defined.
- Disaster recovery reasoning dataset
- COM-DR-001
- Manufacturing/OT workflow dataset
- COM-MFG-001
- Defensive cybersecurity dataset
- COM-CYBER-001
- Enterprise architecture decision dataset
- COM-ARCH-001
Expertise behind the datasets
Comans Services is based in Melbourne, Australia. Our team works across applications, automation, networks, infrastructure, recovery and ongoing technology support.
That operational experience informs the data we acquire or create, the problems we model and the quality checks we design. Our technology project stories show the work behind that experience.
See our technology projectsKnow where each record comes from
We’re developing a review process for acquired and Comans-created material, covering provenance, rights, privacy and technical quality before commercial release. Each dataset needs a documented basis for its intended sale or licensing and use in model development.
- Origin and rights
Record where material came from, how it was acquired or created, and which rights permit its use, modification and sale or licensing.
- Privacy and confidentiality
Review the material for personal information, credentials, confidential details and restricted third-party content.
- Technical quality and release history
Document quality checks, provenance and review evidence. Define each dataset’s identifier, version, coverage, limitations, refresh history and licence terms before release.
A few questions you may have
What is the difference between training data and evaluation data?
Training data is used to develop or adapt a model’s behaviour. Evaluation data is used to assess how a system performs. A reliable evaluation requires attention to whether the tested system has already been exposed to the tasks or their answers.
Can we get an evaluation sample now?
COM-ITOPS-001 is in development. Contact us to discuss your dataset requirements. We’ll confirm sample availability, purchase or licensing options and access terms as the first release develops.
Does this use your clients’ information?
Access to client systems does not give us permission to use their information for AI training. Acquired material and any proposed use of client-derived information need a separate review of rights, privacy and confidentiality, with appropriate permissions for the intended use.
What will a dataset licence cover?
Each offer will identify the dataset and version, coverage, delivery format, permitted uses, price and any access to refreshed or new releases. Buyers can discuss a single dataset, regular releases or a data category they need. Availability and terms will be confirmed before purchase.
What data does your model need next?
Tell us which dataset or data category you need, how you intend to use it and whether you’re interested in one release or regular supply. We’ll discuss the catalogue’s development and future licensing options.
Please describe the work at a high level. Keep client records, personal information and credentials out of the initial enquiry.