Takes someone who doesn't code from “I want to predict this” to a trained model, explained and serving predictions, in Spanish, without the data ever leaving their computer.
From the goal to the algorithm, with a reason
Confusion matrix · the winner
On screen
Real screenshots of the application
What it is
Someone with a data problem — fraud, demand, customer churn — who is not a data scientist had two options: pay a consultancy, or learn scikit-learn alone, which means knowing which algorithm to pick, how to prepare the data and what the metrics mean. Cloud AutoML platforms solve part of it, but they ask you to upload your data to a third party and hand back numbers without explanation. iPrepara puts the whole cycle — profile, recommend, train, explain, predict, export and watch for drift — into a single local application, and explains every result in plain language.
Who it's for
A professional at a small company, or a student, with a spreadsheet and a business question — someone who knows their field but doesn't code, and who wants to understand what they get, not just get it.
How it works
From the goal to the algorithm, with a reason
You choose what you want to achieve, upload or pick a dataset, and the app profiles it on its own and ranks the candidate algorithms, justifying each position. It does not hand back a black box: it hands back a reasoned ranking.
Real training, not simulated
Experiments really train with scikit-learn, up to four at once with live progress, signed with HMAC. They return metrics, a confusion matrix, feature importance and optional cross-validation.
Data study and repair
It finds duplicates, nulls, numbers stored as text, unnormalised dates and outliers, and proposes the repair with 21 preparation tools. A copy of the original is always kept.
Predict, export and watch
One-off predictions, CSV batches or a REST endpoint. The model exports as a .pkl with its Python code, produces an HTML report, and is monitored for drift.
A teaching library with live simulation
52 algorithm cards and 113 use cases running from problem to algorithm, each with a simulation over plausible synthetic data. From any of them you can jump to “try it with MY data”.
Scope
Who it is designed for:
- For one person on their own computer: it listens only on loopback and protects its database from a second server.
- It focuses on classical machine learning; LSTM, BERT or CNN appear in the library as cards with a simulation, to understand how they work.
Frequently asked
Isn’t this a wrapper over scikit-learn that anyone could build in a Jupyter notebook?
Whoever codes already has their notebook; iPrepara is built for someone who has the data and the business question but would never write a line of Python: it explains in plain language what every figure means, diagnoses the data and proposes a repair you confirm before it is applied, and nothing leaves your computer. Cloud AutoML aims at another scale; iPrepara aims at not needing to know what a Jupyter notebook is.
Version 1.0.0 is closed and passed its 851 tests. A “v2” evolution is under way, switched off by default until it is complete.