Runs in your browser

Train models on your own data, without it leaving your browser.

Drop a dataset, pick what to predict, and click Train Models. ML Lab does the work on your own machine and shows you the results in a few seconds.

ml_lab · leaderboard
1Gradient Boosting best0.91
2Random Forest0.88
3Logistic Regression0.82
4Naive baseline0.61
example run · held out 20% to test · best model marked

How it works

Four steps from a file to a result.

There is no setup and no code to write. You stay in the browser the whole time.

01

Drop a dataset

Drag in a CSV or one of the other supported files. It is read right here, nothing is sent anywhere.

02

Pick a target

Choose the column you want to predict, or let ML Lab choose. It also suggests which columns to keep.

03

Train models

One click. ML Lab works out the task type, trains a small set of models, and holds back part of the data to test them.

04

Read the results

See which model did best, how it compares to a naive baseline, the plots, the top features, and a plain read.

Privacy

Your data never leaves the device.

The file you drop in is read and trained on inside the browser. It is not uploaded, and nothing about its contents reaches a server. There is no copy on our side because there is no transfer in the first place.

It stays on your machine

Reading the file, picking columns, and training all happen locally. The data is gone from memory when you close the tab.

You can check it yourself

Open the browser network tab and start a training run. You will see no requests go out. That is the proof, not a promise.

No account to try it

You can run on your own data without signing in. Nothing is logged about the dataset because nothing is sent.

browser devtools
ElementsConsoleNetworkSources
NameStatusTypeSize
0 requests during training
recording network activity, the list stays empty
0 requests · 0 B transferredrecording

Specs and limits

What it reads, and where it stops.

ML Lab runs on the CPU in your browser, so it is built for small to mid sized datasets. Here is the honest envelope.

Supported files

CSVTSVTXTJSONJSONLNDJSON

Drop tabular data in any of these. ML Lab reads the columns and recommends which ones to keep before you train.

Limits

ComputeCPU only
File size10 MB
Rows10,000
Columns200
Model features80

What it trains

ML Lab detects the task from your target column, then trains a small set of models and holds back part of the data to test them.

Classification predict a category
Regression predict a number
Every run is scored against a naive baseline
On the roadmap

More updates coming soon.

We're working on handling bigger files and more kinds of models, and making training quicker. It'll still run right in your browser.

GPU accelerationLarger datasetsMore model typesModel exportShareable reports

What you get

Run it once and read the whole picture.

Pick a sample dataset below and train it. This demo runs the same kind of output ML Lab gives you: a leaderboard, metrics against a baseline, plots, the features that mattered, and a plain read.

ml_lab · new rundemo
Drop a dataset
or pick a sample below
Sample datasets

Pick a sample and click Train Models. The leaderboard, plots, and a plain read will land here.

Bring a dataset. Keep it on your machine.

Drop a file, pick what to predict, and read the results in a few seconds. Nothing leaves the browser.