Listing Thumbnail

    Model Training and Fine-Tuning on AWS - Non-HCLS

     Info
    Adapt and evaluate an existing AI model, or build the agreed training infrastructure and pipelines on AWS, for workloads outside healthcare and life sciences. Define the data, model, compute budget, baseline and acceptance tests before delivery.

    Overview

    Cloud Combinator helps teams outside healthcare and life sciences implement an agreed model adaptation, evaluation or training-engineering workload on AWS. The customer retains ownership of the business objective, scientific or model decisions, approved data and release decision. This service provides the engineering work and evidence needed to assess the selected approach.

    Choose one or both delivery tracks in the statement of work:

    Model adaptation and evaluation

    • Review the intended use, authorised datasets, model licence and baseline measures.
    • Prepare the agreed training and evaluation datasets, with documented versions and separation of training and evaluation data.
    • Configure and run the agreed fine-tuning or adaptation process for an existing model.
    • Evaluate the baseline and adapted model on representative tests, record limitations and compare the observed results against agreed criteria.
    • Hand over the agreed model artefacts, configuration, evaluation workflow and results, subject to the applicable model and data licences.

    Training infrastructure and pipelines

    • Agree the workload, approved model and training plan, compute requirements, cost controls and acceptance tests.
    • Configure the selected AWS training environment and access controls. Amazon SageMaker or other approved AWS compute services may be used where suitable.
    • Implement the agreed data and training pipelines, experiment records, checkpoints, monitoring and reproducibility measures.
    • Run the agreed pipeline or training validation work and document the observed results.
    • Hand over the agreed source, infrastructure configuration, operating procedures and recovery guidance.

    Prerequisites and acceptance

    The customer provides an AWS account or approved deployment environment, authorised datasets and data rights, a named model or scientific owner, an approved model and licence, a compute budget and evaluation criteria. Model availability, supported adaptation methods, processing region, data handling and output ownership are confirmed during scoping. The statement of work specifies which preparation, adaptation, training, evaluation and deployment tasks are included.

    Acceptance is based on the agreed engineering deliverables and test evidence. Where model-quality thresholds are acceptance criteria, the offer defines the dataset, measures, evaluation method and what happens if those thresholds are not met. A completed training run does not establish production suitability. The customer reviews the evidence and approves any release. No model-performance improvement, research result or production approval is guaranteed.

    Scope boundaries

    This package excludes healthcare and life-sciences workloads, which use the separate HCLS service packages. Training a new foundation model from scratch, scientific research ownership, clinical or diagnostic validation, regulatory certification, general application development and ongoing managed operations are outside this package. Deployment of model endpoints is included only when expressly scoped. AWS infrastructure, training compute, storage, model usage and third-party software or licence charges are separate.

    Scope, delivery dates, staffing, acceptance tests, support hours, the support period, refund terms and service price are agreed in the statement of work and private offer before delivery starts. Charges use the selected AWS Marketplace service dimension. Any AWS funding is subject to eligibility and approval and is confirmed separately.

    Highlights

    • Select model adaptation and evaluation, training infrastructure and pipelines, or an agreed combination for a non-HCLS workload.
    • Define data rights, model licences, compute controls and evaluation criteria before training starts.
    • Receive the agreed engineering outputs, reproducible configuration, test evidence and operating handover.

    Details

    Delivery method

    Deployed on AWS

    Pricing

    Custom pricing options

    Pricing is based on your specific requirements and eligibility. To get a custom quote for your needs, request a private offer.

    How can we make this page better?

    Tell us how we can improve this page, or report an issue with this product.
    Tell us how we can improve this page, or report an issue with this product.

    Legal

    Content disclaimer

    Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.

    Support

    Vendor support

    Contact Cloud Combinator at hello@cloudcombinator.ai  or +44 20 3588 2584.

    Before purchase, Cloud Combinator helps you assess suitability, scope and prerequisites. We respond to enquiries received through the listing within two business days.

    After private offer acceptance, we contact you or provide next steps within two days, unless the offer specifies another arrangement. During delivery, we provide configuration guidance, troubleshooting and handover support for the agreed scope. Support hours, the post-handover support period and any response targets are specified in the statement of work and private offer. Use the support contact shown on this listing and the fulfilment page for service issues and refund requests; refund terms are set out in your offer.