ProductOverviewHow It WorksCapabilitiesFor Your Team
TrustOverviewData Protection & GovernanceSecurity & AccessOwnership & ConfidentialityAudit & Accountability
ResourcesTechnical OverviewAPI OverviewSupport
CompanyAbout AscentMLContactRed Mountain Innovations
Talk to AscentML

Everything you need to move from data to predictions

AscentML gives your team a clear, connected way to organize data, request custom models, review results, control deployment, and use predictions.

The platform is designed to keep the experience straightforward for business users while making additional depth available to analysts and technical teams when needed.

  1. 1

    Organize work around real business problems

    Projects

    Projects bring related work together in one place.

    Use them to organize:

    • Datasets
    • Training requests
    • Published models
    • Deployments
    • Prediction history
    • Notes and comments

    Projects help your team understand what exists, what changed, and what needs attention next.

    Datasets are not locked to a single project and can be reused across multiple efforts where appropriate.

  2. 2

    Bring your own data

    Versioned datasets

    Upload CSV or Parquet datasets through a guided workflow.

    AscentML keeps dataset versions immutable so your team can maintain a clear history of what was used and when.

    Datasets can be:

    • Versioned
    • Reused across projects
    • Connected to related models
    • Soft deleted and recovered when needed

    Validation and screening

    Before data moves into model development, AscentML can identify issues such as:

    • Missing values
    • Constant or near-constant columns
    • Potential target leakage
    • Possible sensitive or prohibited information

    Customers can identify target fields and known leakage fields as part of the upload process.

    Ambiguous sensitive-data findings can be quarantined for review before the dataset proceeds.

    Dataset upload — versioned, validated, and screened before use.
  3. 3

    Request a model without designing a training pipeline

    Managed training requests

    Start with the business objective, not the algorithm.

    Select the project and dataset, identify what you want to predict, and provide the context behind the request.

    AscentML manages the advanced training process internally, including model development, experimentation, optimization, and packaging.

    Your team can follow the request as it moves through review, training, validation, and publication without managing the internal training machinery.

  4. 4

    Review models in terms that matter

    Published models

    When a model is ready for customer review, AscentML publishes an immutable model version into your environment.

    Customers can review information such as:

    • Model status
    • Performance
    • Dataset lineage
    • Feature importance
    • Intended use
    • Release notes
    • Limitations
    • Comments
    • Deployment history
    • Prediction history

    Business users can remain focused on the overall outcome, while advanced users can access richer metrics and diagnostic information.

    Explainability

    AscentML supports multiple levels of model explanation.

    That can include:

    • Plain-language primary drivers
    • Contribution views
    • Detailed SHAP-based explanations for advanced users

    Per-prediction explanations can be generated on demand against the exact published model version that produced the result.

    Model review — performance, lineage, and explanations in one place.
  5. 5

    Control what gets deployed

    Model deployments

    Your team decides which published model versions are put into use.

    Capabilities include:

    • Multiple model versions
    • Manual promotion
    • Manual rollback
    • Percentage-based traffic allocation

    Where comparison is useful, traffic can be divided between model versions—for example, 80% to one version and 20% to another.

    Routing is intentionally kept simple in the initial product. Rules-based routing, scheduled deployment changes, and automatic promotion or rollback are not part of the current MVP scope.

  6. 6

    Put predictions into your workflow

    Batch predictions

    For business workflows, users can run predictions against uploaded files through a guided process.

    The workflow includes:

    1. Select the project and model
    2. Upload the inference dataset
    3. Validate the input schema
    4. Submit the job
    5. Track progress
    6. Download the results

    Jobs can move through states such as validating, queued, running, completed, completed with warnings or errors, failed, or cancelled.

    API inference

    Technical teams can integrate published models directly into applications using API access.

    API capabilities include:

    • User-scoped API keys
    • Key creation, rotation, and revocation
    • Per-key quotas
    • Per-key analytics
    • Optional project or model restrictions
    • OpenAPI / Swagger documentation

    API keys cannot grant more access than the user who owns them.

  7. 7

    See what produced each prediction

    Prediction outputs are designed to preserve context.

    Depending on the model and workflow, results can include:

    • Prediction
    • Confidence
    • Model version
    • Timestamp
    • Request information
    • Optional explanation

    This helps teams understand not only the result, but which published model produced it.

  8. 8

    Give each user the right level of complexity

    Standard mode

    For business users who want clear workflows and outcomes.

    The experience emphasizes:

    • Guided language
    • Reduced information density
    • Recommended next actions
    • Simplified model interpretation

    Advanced mode

    For analysts and technical operations users who need more detail.

    Additional information can include:

    • Model metrics
    • Dataset lineage
    • Diagnostics
    • API analytics
    • Audit detail
    • Deployment controls

    Permissions determine what a user can do.

    Experience mode determines how much complexity they see while doing it.

  9. 9

    Govern access and activity

    Users and permissions

    AscentML supports customer-specific roles and permission-based access control.

    Customer administrators can manage users within the access boundaries established for their organization.

    Audit history

    Important state changes are recorded so teams can understand what happened, when, and by whom.

    Audit coverage includes areas such as:

    • User administration
    • Dataset activity
    • Model deployment
    • API key operations
    • Quota changes
    • Privileged administrative actions

    Notifications and comments

    Users can receive in-app and email notifications related to relevant platform activity.

    Comments and notes can be attached to projects, datasets, models, deployments, and training requests, with support for customer-visible and internal-only communication.

Built around connected history

AscentML is designed so datasets, models, deployments, and predictions remain connected rather than becoming isolated records.

That means your team can trace:

The data that was used → the model that was published → the deployment that was approved → the prediction that was produced.

That history supports better decisions, clearer accountability, and easier review.

Focus on the outcome. Go deeper when you need to.

Business teams can use AscentML without operating a machine learning workbench.

Analysts and technical teams can access additional metrics, APIs, lineage, explanations, deployment controls, and operational detail when required.

For implementation-level information, architecture, and API behavior:

Ready to put your data to work?

Start with a problem you would like to predict more effectively. AscentML provides the managed development process and the platform needed to turn the result into something your team can actually use.