Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

FEDOT

fedot.readthedocs.io

  • Free plan
  • Windows
  • macOS
  • Linux
  • API
  • Self-hosted

What FEDOT is

FEDOT is a free, open-source framework for automatically generating composite models from data. It supports classification, regression, clustering and time-series forecasting, working with tabular, text and image data, including combinations of these types. The workflow covers preprocessing, model selection, tuning, cross-validation and serialization. Users can omit parameters for greater automation or supply them to guide pipeline composition. FEDOT uses GOLEM to optimize graph-based pipelines with meta-heuristic methods, and offers presets including best_quality, fast_train, stable, gpu, ts and automl. Inputs can come from CSV files, pandas DataFrames, NumPy arrays and time-series CSV data. Its API can be called from a console without Python code, with predictions saved as CSV files. GPU evaluation uses RAPIDS and supports a specified set of models, including Ridge, Lasso, LogisticRegression, RandomForestClassifier, RandomForestRegressor, KMeans and SVC. Installation is available through pip, with optional image, text-processing and DNN dependencies. FEDOT is maintained by NSS Lab at ITMO University's National Center for Cognitive Technologies and runs on Windows, Linux and macOS.

Who it suits

FEDOT suits developers and researchers who want an open-source AutoML framework for supported classification, regression, clustering or forecasting tasks. It offers both automatic and parameter-guided pipeline composition.

Good to know, and what to watch

  • Free and open source under the 3-Clause BSD license.
  • Works with tabular, text and image data.
  • Supports automated or parameter-guided pipeline composition.
  • Can be called from a console without Python code.
  • Runs on Windows, Linux and macOS.
  • GPU evaluation supports only a listed set of models.
  • Image, text-processing and DNN dependencies are optional extras.
  • Workflow interface is code-based.

FEDOT

Free

Open source

Getting started

  1. Open the FEDOT documentation website
  2. Install the package with pip using `pip install fedot`
  3. Add `fedot[extra]` for optional image, text-processing, and DNN dependencies
  4. Prepare data from CSV, pandas, NumPy, or time-series CSV sources
  5. Choose a preset or provide parameters for pipeline composition
  6. Run the API in Python or call it from a console

Limits to check

FEDOT is offered as a free free open-source plan. GPU evaluation is limited to Ridge, Lasso, LogisticRegression, RandomForestClassifier, RandomForestRegressor, KMeans, and SVC.

The Club's review

FEDOT is a free, open-source framework for building machine-learning pipelines in code. It is aimed at developers, data scientists and researchers who want AutoML assistance without giving up control of pipeline design. The trade-off is a technically oriented workflow and narrowly scoped GPU acceleration, rather than a polished visual tool.

For teams comfortable with Python and self-hosting, FEDOT offers unusually broad task and data coverage at no license cost. Readers who need a managed service, graphical interface or broad GPU model support should look elsewhere.

Overview

FEDOT addresses binary and multiclass classification, regression, clustering, and univariate or multivariate time-series forecasting. It is distributed under the 3-Clause BSD license, so it can be used in projects and research, including commercial work. The project is developed and maintained by the NSS Lab within the National Center for Cognitive Technologies at ITMO University in Russia.

This is a code-first framework rather than a point-and-click application. Its strongest fit is a practitioner who wants automated model selection and feature engineering while retaining the ability to inspect, constrain or compose pipelines. It is less appropriate for business users who need a visual workflow or a hosted environment.

Key features

FEDOT covers preprocessing, model selection, tuning, cross-validation and serialization. InputData can be created from CSV files, pandas DataFrames, NumPy arrays and time-series CSV files. Tabular, image and text data are supported, including multimodal combinations from more than one source. Preprocessing addresses infinite and missing values, binary and non-binary categorical features, and stray spaces in categorical data, reducing several common cleanup steps before modeling.

Automation can be tuned to the job. Omitting parameters enables full automation; supplying them allows partial automation and manual pipeline composition. Presets include best_quality, fast_train, stable, auto, gpu, ts and automl, with auto as the default. That range supports quick baselines as well as more deliberate experimentation, but making good choices still requires familiarity with machine-learning workflows.

GOLEM supplies optimization and learning for graph-based pipelines using meta-heuristic methods. FEDOT’s model collection is drawn mostly from scikit-learn, statsmodels and Keras; the project also describes integration with scikit-learn, CatBoost and XGBoost and support for custom libraries. The default validation uses five-fold cross-validation, and additional metrics can be supplied to the optimizer to help address bias.

Installation is straightforward through pip install fedot. The optional fedot[extra] package adds image, text-processing and DNN dependencies. The API can also be called from a console without writing Python, with predictions saved as CSV files. GPU evaluation relies on RAPIDS and currently covers Ridge, Lasso, LogisticRegression, RandomForestClassifier, RandomForestRegressor, KMeans and SVC only. GPU-oriented workloads outside that set will not receive the same acceleration.

Pricing

PlanPriceWhat you get
FEDOTFreeOpen-source AutoML framework under the BSD 3-Clause license

There is no paid tier, free trial or subscription renewal. The free plan is the complete self-hosted framework, making it suitable for individual experiments, research and teams able to operate their own environments. The price does not include a hosted service or a graphical workflow; infrastructure, setup and ongoing maintenance remain the user’s responsibility.

Platforms

FEDOT supports Windows, Linux and macOS, and is available as an API for self-hosted deployments. Its console access is useful for scripts and batch jobs, while the Python installation suits teams building it into existing codebases.

Who it's for

Choose FEDOT if you need one open-source toolkit for tabular, text, image or mixed data and want to decide how much pipeline construction is automated. It fits research groups and engineering teams that value extensibility, reproducible code and a permissive license.

Look elsewhere if non-technical stakeholders must build models through a browser, if you want a vendor-managed runtime, or if your intended GPU models are outside FEDOT’s supported RAPIDS set.

Pros and cons

Pros

  • Broad problem coverage: classification, regression, clustering and both univariate and multivariate forecasting are handled in one framework.
  • Multimodal inputs: tabular, text and image data can be combined, with several common input formats accepted.
  • Adjustable automation: users can move from fully automated searches to constrained or manually composed pipelines.
  • Open and extensible: the 3-Clause BSD license and stated support for custom libraries suit research and production integration.
  • Lifecycle support: preprocessing, tuning, validation and serialization reduce the need to assemble separate components.

Cons

  • Code-first operation: there is no graphical workflow for users who prefer visual model building.
  • GPU limitations: RAPIDS evaluation is restricted to seven named model families, limiting acceleration choices.
  • Self-hosted responsibility: users must handle installation, runtime infrastructure and operational upkeep themselves.
  • Technical learning curve: effective preset and parameter choices require machine-learning knowledge even when automation is enabled.

Alternatives

BigML is the better choice when a web-accessible platform, API and freemium entry point are more important than a wholly open-source framework; its free plan allows unlimited tasks and storage but caps each task at 16 MB, two parallel tasks and one user.

JADBio suits teams wanting a hosted workflow with seats, projects, storage and model-export allowances, with paid team pricing available.

Amazon SageMaker Autopilot is preferable when managed cloud operation and pay-as-you-go billing are the priority.

AutoKeras is a free pip-installed Python package for users focused on an alternative open-source AutoML approach.

LightAutoML is another free, self-hosted Python library, released under Apache 2.0.

MLJAR AutoML offers a free MIT-licensed Python package alongside a separate hosted free tier.

Auto-PyTorch is a free BSD-licensed option for Linux and self-hosted deployments.

AutoGluon is a free Apache 2.0 Python library supporting Linux, macOS, Windows and self-hosting.

For a wider comparison, see AutoML Software.

Verdict

FEDOT is a strong choice for developers and researchers who want broad, open-source AutoML with adjustable pipeline automation, multimodal data handling and full self-hosting. Choose it for control and breadth at zero license cost; choose another platform when you need a visual or managed experience, or GPU evaluation beyond its supported model list.

The verdict

FEDOT covers several stages of the machine-learning workflow and supports multiple data types and tasks. Users can choose how much pipeline composition to automate, while GPU evaluation is limited to the specified models.

Questions people ask

Is FEDOT free?

Yes. The FEDOT plan is free, and the project is open source under the 3-Clause BSD license.

Which operating systems does FEDOT support?

The quick-start guide lists Windows, Linux, and macOS. Self-hosted deployment is supported.

Can FEDOT work with images and text?

Yes. It supports tabular, image, text, and multimodal data.

Can I control the level of automation?

Yes. Omitting parameters enables fuller automation, while supplied parameters guide manual pipeline composition.

How can FEDOT be installed?

Install it with pip using `pip install fedot`; optional dependencies are available through `fedot[extra]`.

What the maker publishes

automation
Users can choose full automation by omitting parameters or partial automation by supplying parameters for manual composing. source
Automation controls
Users can adjust automation by omitting parameters for full automation or supplying parameters for partial automation. source
cli
Its API can be called from a console without Python code, and predictions are saved as CSV files. source
Contributions
The project welcomes contributors to report bugs or propose enhancements through its GitHub issues. source
data_inputs
InputData can be created from CSV files, pandas DataFrames, NumPy arrays and time-series CSV data. source
Data types
FEDOT works with tabular, image, and text data, including multimodal data from more than one source. source
Extensibility
The project says FEDOT supports widely used ML libraries such as scikit-learn, CatBoost, and XGBoost, and allows custom libraries to be integrated. source
gpu
GPU evaluation uses RAPIDS and currently supports Ridge, Lasso, LogisticRegression, RandomForestClassifier, RandomForestRegressor, KMeans and SVC. source
installation
FEDOT can be installed with pip using `pip install fedot`, with optional image, text-processing and DNN dependencies available through `fedot[extra]`. source
Installation
The project README documents installation with pip, including an extra option for image and text processing and deep neural networks. source
license
FEDOT is published under the BSD-3 license for use in projects and research. source
Maintainer
FEDOT is developed and maintained by the NSS Lab team, part of the National Center for Cognitive Technologies at ITMO University in Russia. source
maker
FEDOT is developed and maintained by the NSS Lab, part of the National Center for Cognitive Technologies at ITMO University in Russia. source
ML lifecycle
FEDOT covers preprocessing, model selection, tuning, cross-validation, and serialization. source
Model libraries
FEDOT uses models mostly from scikit-learn, statsmodels, and Keras. source
model_presets
The framework provides presets including best_quality, fast_train, stable, auto, gpu, ts and automl, with auto as the default. source
multimodal_data
FEDOT can work with multimodal data including tables, texts and images. source
Operating systems
The quick-start guide lists Windows, Linux, and macOS as supported operating systems. source
pipeline_optimization
FEDOT uses the open-source GOLEM library for optimization and learning of graph-based pipelines with meta-heuristic methods. source
Pipeline optimization
FEDOT uses the GOLEM library for optimization and learning of graph-based pipelines with meta-heuristic methods. source
preprocessing
Its preprocessing handles infinite values, missing values, binary and non-binary categorical features, and extra spaces in categorical data. source
Preprocessing
FEDOT can replace infinite values, drop rows with many missing values, binarize binary categorical data, and trim extra spaces in categorical features. source
purpose
FEDOT is an AutoML-like framework for automated generation of data-driven composite models. source
Purpose
FEDOT is an open-source framework for automated modeling and machine learning that builds end-to-end solutions using an evolutionary approach. source

Checked against the maker's pages on 2 Oct 2026.