Workspace / Dashboard
Docs GitHub Connecting
Try the pipeline in your browser

Up to 1,000 rows / 2 MiB. Files stay in this tab. Reloading clears your data and results. Local rules and byte-bigram training only — no JEV calls or LLM training.

Self-host the full version ↗
JEV DATAOPS / PUBLIC BROWSER DEMO

JEV DataOps Browser demo.

Bring your domain data. Screen for quality. Train and evaluate your next model.

Explore the domain guide
Datasets0

Currently loaded datasets

Data records0

Rows across loaded datasets

Completed runs0

0 recent runs loaded

01
UploadJSONL / CSV
02
ScreenKeep · Review · Reject
03
Data evalMeasure data quality
04
TrainByte-bigram
05
Model evalValidate on held-out data
01

Prepare your data

Upload a dataset or try your first experiment.

Drop your dataset here

or browse files

JSONL / CSV · Streamed in batches

Supports text, instruction–response pairs, and conversations. See Docs for formats.

Your data preview will appear hereNo dataset yet? Select Use example to try it out.
02

Configure your workflow

Choose how to screen, train, and evaluate.

CONFIGURATION

Demo uses local rules. Connect JEV to apply basic screening for your selected domain.

Confidence, concurrency and request limit apply to JEV screening only.

i

Demo modeLocal rules and pipeline validation only. No JEV calls or LLM training.

Add a dataset to get started

Runs & insights

Follow every step, from screening decisions to evaluation results.

0 RUNS
EVERY EXPERIMENT, A STEP FORWARD

Ready to put your data to work?

Start a run to follow screening results, training progress,
and model performance on held-out data.

Connection settings

Enter the access token configured on your server. It is stored only for this browser tab session.

Configure JEV API keys and model paths using server environment variables.