Skip to content
Predict.ai
Serving

Serve predictions in production.

Promote any completed training — from a goal's tournament or a manual pipeline — to a hosted endpoint. Run inference against the freshest data, and watch latency, drift, and quota in one place.

Every inference across every deployment — latency, horizon, and the forecast itself.

Deployments

Hosted endpoints. Autoscaled. Rollback-ready.

Promote an eligible completed training — manually or on a schedule — to a stable endpoint. Pick shared hosting for cost efficiency or dedicated hosting for isolated capacity.
  • Shared or dedicated hosting — same API surface
  • Autoscale by traffic, route by region
  • Auto-promote on a schedule, or with one click
  • Atomic rollback to any prior promotion
Promotion lifecycle, usage, and identifiers — together, with a one-click API guide.

Inference

Request a forecast with one API call.

An inference is a single request to a deployed model: your application calls the endpoint, the model runs on the latest data from its segment, and the response returns the forecast with quantile bands. Use REST for one-off calls, streaming for continuous updates, and batch for backfills.
  • REST, streaming, and batch — same call signature
  • Inputs come from the model's segment, so the data shape never drifts
  • Per-call latency, drift, and cost — captured by default
A single inference — quantile bands, model quality, and every forecast step.

Promote in one click

From training to a stable endpoint — auto-promote on a schedule or manually with a click.

Shared or dedicated

Cost-efficient sharing for many small models, dedicated isolation for the heavy ones.

Regional data residency

Run workloads in the workspace's home region and keep customer content at rest there.

Live observability

p50/p95/p99 latency, error rates, drift, and quota — across every deployment.

REST · streaming · batch

One interface. One call signature. Three workloads.

Per-call cost

Every prediction is metered. Quotas, alerts, and credit-aware billing built in.

Reversible

Every promotion has a rollback target. No fire drills, no late-night rollouts.

Access and audit controls

Workspace isolation, RBAC, SSO, scoped API tokens, audit logs, and regional residency.

Put your data to work.

Connect your own data or start with a sample. Explore automatic insights, ask questions, engineer features, and build a forecast in the same workspace.

Start with your dataTalk to sales