One platform to generate, test, and evaluate your synthetic time-series data

Rockfish is a modular platform for generating, testing, and evaluating synthetic time-series data. Each module does one job — and they compose into end-to-end workflows for ML training and agent evaluation.
‍

The Rockfish Workflow

You generate data or train a model, check its quality, then branch into two downstream tracks depending on what you need.
‍

Input

Data or Schema

Production time-series, a schema definition, or plain-language intent

Step 1

Rockfish Platfom

Train Rockfish Model or Generae Data with DataFuel or SchemaFuel

Step 2

Eval Studio

Measure data fidelity, privacy, and agent correctness — with actionable scores

Output

Model or Data

Trained custom Model or Baseline Data ready for downstream use

Then choose your track

Track A - Generate More Data

Scenario Studio

Takes your model or dataset and generates scenario-specific variations — injecting edge cases, rare events, and incident patterns, blending to expand your training data.

Track B - Evaluate Agents

Scenario Studio + AgentFuel

Scenario Studio injects patterns into your data, then AgentFuel generates prompts, queries, and expected responses from those scenarios ready for AgentEval to score.