From raw data to an evaluated model
Three ways to work together, depending on where you're starting. Each level builds on the one before: buy prepared data, commission a custom data-to-model project, or keep a model improving month after month.
Buy a dataset
A prepared dataset for a specific training or fine-tuning objective. Optionally, I train a model on it for you.
Best if: you need data, not a full engagement.
Custom data → model
You bring raw or existing data. I turn it into a training-ready dataset, train on it, and evaluate the result.
Best if: you have data and a target model.
Managed monthly
A recurring agreement for continuous preparation, training, evaluation and updated model versions.
Best if: your model needs to keep improving.
Ready-to-use training datasets
The entry point: prepared training data without paying for a full training engagement. Buy individual datasets, or subscribe to recurring deliveries where available.
$0.45 per row · $1,000 minimum · $0.30/row above 10,000 rows · monthly updates $300/mo · non-exclusive licence (exclusive on request)
Catalogue in progress. Datasets are currently prepared on request. Tell me your training objective and I'll say what's available or what it would take to build. Datasets are only resold to multiple clients where licensing, ownership, privacy and usage rights allow it.
- Cleaned and structured training data
- Normalisation and deduplication
- Labelling and organisation where required
- Quality checks and validation
- Formatting for compatible training workflows
- Dataset documentation and usage guidance
Dataset + model training
After buying a dataset, I run a compatible training or fine-tuning workflow on it, evaluate the results and deliver an updated model version. This is priced above the dataset alone because it adds engineering, training and evaluation work.
Custom data preparation + model training
An end-to-end project for clients who already have data. It costs more than buying an existing dataset because it includes data engineering specific to you, quality assurance, training and evaluation.
Data preparation alone, without training, typically starts around $400–$1,000 depending on scope.
Start a project- Data assessmentReview your data, target model and training objectives.
- Automated data preparationMy in-development data-prep system assists with cleaning and structuring messy or inconsistent data.
- Human quality assuranceI review the output for incorrect deletions, leftover duplicates, formatting problems and inconsistencies.
- Dataset validationConfirm it meets the agreed requirements and suits the training workflow.
- Training or fine-tuningRun supported jobs on the agreed model or provider.
- EvaluationTest against agreed benchmarks, and against a baseline where feasible.
- DeliveryThe model version or supported deployment, with evaluation results and documentation.
Managed monthly model training
For clients who want ongoing model improvement instead of buying projects one at a time. You get a defined base fee covering a defined workload, at a better effective rate than equivalent standalone work.
- Limited data-processing volume
- Basic cleaning, structuring and human review
- 1–2 scheduled training runs
- Basic model evaluation
- Updated model + summary report
- Higher data-processing volumes
- Repeated preparation and review cycles
- Multiple training or fine-tuning runs
- Benchmarking against previous versions
- Detailed performance reporting
- Large or complex datasets
- Extensive preparation and validation
- Multiple iterations and model versions
- Expanded evaluation and regression testing
- Faster delivery, more reporting and support
Not sure which tier fits?
Tell me your target model, what data you have (or need), and your goal. I'll recommend a starting point.
Request a quote