Pradra StudioDocs
Sign in

How the platform works

Pradra Studio takes an AI capability through one continuous life: define the problem, prepare the data, train and evaluate, deploy, and keep watching it in production. This page is the map; every stage links to its own section.

The lifecycle at a glance

The arrows only move forward through a quality gate: a trained model reaches production only when it passes the acceptance criteria you defined. And the loop closes — when monitoring sees the real world drift away from the training data, retraining feeds a fresh candidate back through the very same gate.

Stage by stage

  1. Define the solution. Create an AI Solution for the business problem, then add one capability per prediction you need. Each capability declares its contract — the inputs it takes and the outputs it returns.
  2. Prepare the data. Upload a dataset (or pull one in through a connector), let the platform validate it, and define the feature set the model will learn from.
  3. Train and evaluate. Submit a training run. Every run is recorded as an experiment; the resulting candidate model is evaluated against your acceptance criteria.
  4. Deploy. Promote the model that passed and deploy it. Your applications call a stable endpoint — in real time or in batch.
  5. Operate. Monitoring tracks health and quality, drift detection watches the data, and budgets keep generative spend under control.

The same rails for every kind of AI

Classic machine learning, generative capabilities built on large language models, and agents all live inside the same solutions, pass the same kind of quality gate, and are monitored in the same place. Your team learns one workflow, not three.

Everything is versioned

Datasets, feature sets, models, prompts, and knowledge sources are all versioned: a change never overwrites what production depends on, and going back is a pointer switch, not a rebuild. That is what makes rollback instant and audits possible — see Governance & audit.

Next