For iPhone & iPad

Quick ML

A complete data-science studio in your pocket.

Free to download. One purchase unlocks everything, forever. No subscription.

Import real datasets, explore and clean them properly, train machine-learning models, and make predictions - the whole workflow, entirely on device. Built for data scientists and students of data science who don’t always have a laptop with Python to hand.

Data Health Check suggesting cleaning steps, each with its reason

Dashboard of stat tiles and charts built from a dataset

Prediction form with sliders and dropdowns for each model input

1,000,000+

row datasets stream on device

machine-learning algorithms - 4 free, all with Pro

statistical tests, plain-English verdicts

accounts, uploads or trackers

Learning data science?

Every screen explains what the numbers mean and why each step matters. See the Python equivalent of what you just did, and export the whole workflow as a notebook with Pro.

See how it shows its working

Already working with data?

Open real datasets, APIs and databases the moment you get them, analyse them wherever you are, then take the models, reports and dashboards with you.

See what it connects to

New to Quick ML? Read the step-by-step guide

No laptop? No problem

Open a dataset the moment you get it

On the train, in a lecture, at a client’s site, on the sofa: import a dataset and have a trained, evaluated model before you’re anywhere near a computer. A million-row CSV is fine - data streams into a local store and never loads into memory at once.

Full descriptive statistics for every column the moment it lands
Histograms, box & violin plots, distribution charts - all expandable to full screen and shareable
Non-destructive: every change is a recorded recipe step, individually undoable and replayed identically when data refreshes - a failed step never touches the data you are looking at

Data screen showing a one-million-row CSV with its columns and export actions — A million-row CSV, comfortable on a phone.

Connectors

Import anything

Real data comes from everywhere. Quick ML meets it there.

CSV, TSV & Parquet

Open delimited files of any size - a million rows streams straight into a local store. Parquet files from your data pipelines open natively.

Excel

Import sheets from .xlsx workbooks.

Word & PDF

Tables inside .docx files and born-digital PDFs. Tap the table you want when a document has several. Merged cells are handled, and multi-page tables are stitched together with repeating headers kept once. Values come from the PDF’s own text, so scanned pages and images of tables are not read.

JSON & XML

Local files are shaped into a table through a visual mapper: tap the table, pick fields, expand lists into rows. The same mapper turns JSON stored inside a text column into new columns.

Web pagesPro

Point at a page, tap the table you want, and it becomes a dataset to clean, train on and refresh later. Multi-row headers and results split across pages are handled. Works with tables in the page itself, not ones loaded by script after it opens.

JSON & XML APIsPro

Fetch a sample, tap the table into shape, and set up paging, custom headers and sign-in once. Every refresh replays the same steps, including from servers on your own network.

GraphQLPro

Post a query and map the reply the same way, cursor paging included.

KagglePro

Browse and download Kaggle datasets with your own API key.

PostgreSQLPro

Query a database directly - credentials stay in your device Keychain.

Image folders & ZIPs

Folders of images grouped by label sub-folder, ready to train a classifier. A ZIP of same-shaped files, such as monthly reports, combines into one table with a source-file column.

Combine projects

Join two projects into one dataset on a shared key column, then join the result again. Combinations nest, and refreshing a project refreshes every source up the chain.

Refresh, append, mergePro

Sources can be re-imported, stacked, or merged on a key as new data arrives. Keep History appends every fetch stamped with its time, so a page that only shows the current table becomes a time series you can chart and forecast.

Explore & clean

EDA that would make your laptop jealous

Profile, visualise, and clean properly - with the reasoning spelled out at every step.

Data Health Check suggesting cleaning steps with reasons — One-tap Data Health Check

Correlation matrix heat map with strongest pairs — Correlation heat maps & pair plots

Data Health Check profiles every column and suggests cleaning steps, each with its reason - swipe away what doesn’t fit, apply the rest in one tap. It names the values it cannot read and points at the tidy that fixes them, and flags extreme values with the value itself
Fills, filters, dedupe, outlier trims, find & replace, type changes - every change a replayable recipe step, like a pandas script you can swipe
Tidy Dates, Numbers and Durations turn messy text into real values: dates stored unambiguously and shown in your own format, grouped digits, percent signs, currency symbols and bracketed negatives read correctly, lap and clock times like 1:23.636 converted to seconds. A text column of numbers is never silently misread
Derived columns: date parts and differences, bins, ratios, lags, rolling windows, combined columns, maths with a constant, and rich text features - word clouds, sentiment, language, readability, people & places, even AI columns generated by Apple’s on-device intelligence
Nine statistical tests - chi-squared, ANOVA, t-test, Mann-Whitney U and more - with plain-English verdicts
One-tap anonymisation: fake names and emails, scrambled phone numbers, postcode areas, dates to month or year - sensitive data handled before it goes anywherePro

Train

Real models, honestly evaluated

Seeded, stratified or time-ordered splits, expanding-window cross-validation for time-ordered data, cross-validated auto-tuning, and full evaluation - R² and RMSE, confusion matrices, ROC and AUC, permutation feature importance, learning curves. Compare multiple models head to head.

Regression

Linear regression
Decision tree
Random forest (auto-tuned)
Boosted trees
Neural network you design visually

Classification

Logistic regression
Decision tree, random forest & boosted trees
Neural network (up to 6 hidden layers)
k-nearest neighbours & Naive Bayes
Linear SVM & Apple NLP text classifier

Clustering

k-means (elbow & silhouette charts)
DBSCAN (density)
Agglomerative (Ward)
Gaussian mixture

Training settings, explained: ridge and lasso regularisation for linear and logistic regression, learning rate, minimum child weight and row and column sampling for boosted trees, sampling for random forests - each described in plain English beside the control. Auto-Tune searches them for you
Honest scores: image classifiers report accuracy on photos held back before training, models that sampled their training rows say so, and a prediction shows the probability for every class
Stop whenever you like: training and Auto-Tune can be cancelled part way, and every long analysis stops when you leave the screen

Linear regression, logistic regression, k-means and the image classifier are free to train. All 19 algorithms, auto-tuning and the neural network designer come with Quick ML Pro.

Trained classification model with accuracy, precision, recall and confusion matrix — Accuracy, precision, recall, F1 and confusion matrix for every model.

How to Do This in Python screen exporting the pipeline as a notebook or script — Your exact pipeline as pandas & scikit-learn code.

Studying data science?

It shows its working

Every screen explains what the numbers mean and why each step matters, from quartiles to overfitting. Then How to Do This in Python shows your exact pipeline as pandas and scikit-learn code, every step with its reasoning - and Quick ML Pro exports it as a ready-to-run Jupyter notebook that fetches the same data source and replays every tidy step. What you did on the sofa becomes the coursework, and the concepts transfer straight to the tools you’re learning.

Working data scientist? Take the results with you

Export any trained model as a Core ML .mlmodel and drop it into an Xcode project Pro
Fast batch predictions over whole files, score held-out sets, compare model versions
Build live dashboards from stat tiles and charts, then export as a PDF Pro
Generate a client-ready PDF report of the entire project in one tap
Share the whole project with other Quick ML users - the import and cleaning recipe travels with it, so they reconnect the source in one tap and get the same cleaned table Pro

Images mode

Machine learning for photos too

Point Quick ML at folders of labelled images and go from raw pictures to a working classifier without leaving your phone.

Understand your classes

Image and class counts, per-class bar charts, and sample browsers for every label.

Average images

See the average image per class - colour, individual channels, or greyscale - and compare two classes side by side.

Train a classifier

Train an image classifier on device, with a true test accuracy measured on photos held back before training, and a confusion matrix.

Classify photos

Classify single photos or whole batches, with the best match and the probability for every class - and, with Quick ML Pro, export the .mlmodel.

Private by architecture

Your data never leaves your hands

Not a promise buried in a policy - a consequence of how the app is built.

Everything computes on device

Importing, cleaning, training, predicting - all of it happens on your iPhone or iPad. Your data never touches a server.

No accounts, no analytics

Nothing to sign up for, nothing tracking you. An easy answer when the dataset is sensitive.

Syncs through your iCloud

Projects sync privately between your own devices using your own iCloud storage. We receive nothing. Rebuildable data is kept out of device backup, so large datasets never count against your iCloud backup space.

Credentials stay in the Keychain

Kaggle keys and database passwords are stored only in your device’s secure Keychain.

The details, in full: Quick ML Privacy Policy

Made for iPad

More screen, same studio

The full workflow with a projects sidebar and room for dashboards to breathe.

Quick ML on iPad showing the projects sidebar and a data dashboard

Pricing

Free to try. One price to own.

The core workflow is free, forever. A single purchase of Quick ML Pro unlocks everything else - no subscription.

Free, for everyone

Import files of any size: CSV, Excel, Word, PDF, parquet, JSON, XML, ZIPs of files, image folders
Explore and clean without limits: statistics, charts, nine statistical tests, the Data Health Check, every tidy and cleaning step, combining two projects on a key
Train the classic algorithm for each task: linear regression, logistic regression, k-means, and the image classifier
Full honest evaluation: confusion matrices, ROC, residuals
Predict, including whole files and held-out test sets
Read your pipeline as Python, export your cleaned CSV, generate PDF reports (dashboards are a Pro feature)
Up to 3 projects, synced privately between your devices

Quick ML Pro

One purchase, yours forever

Live dashboards built from stat tiles and charts
All 19 algorithms, auto-tuning, and the neural network designer
Import from web pages, JSON and XML APIs, GraphQL, Kaggle, and PostgreSQL, with refresh, append, Keep History, and lookup joins
Export models as Core ML and pipelines as Jupyter notebooks
Share whole projects, recipe included, and apply anonymisation
Unlimited projects and full AI column runs

No subscription. No account. Bought Quick ML when it was a paid app? Pro is already yours.

If you have data and a phone,
you have a data-science workstation.

Requires iOS 26 or later. iPhone and iPad.

Want a walkthrough first? Read the how-to guide

Questions? Get in touch