The Origin of dashAI, the No-Code AI Platform for Machine Learning Models Made in Chile
In conversation with FuturIA podcast, the academic from the Department of Computer Science (DCC) and director of the Data and Artificial Intelligence Initiative (IDIA) at Universidad de Chile, also a researcher at the Millennium Institute for Foundational Research on Data (IMFD) and at the National Center for Artificial Intelligence (CENIA), Felipe Bravo, reviewed the origin and development of dashAI, the open-source software that allows users to train artificial intelligence models without the need to code and has already surpassed 31,000 downloads.
FuturIA is a strategic analysis publication on artificial intelligence aimed at executives, leaders, and Spanish-speaking professionals, with an audience in 79 countries. In episode No. 98 of the podcast of the same name, its author, Sergio Vélez Maldonado —consultant and director of the IQAI Institute—, spoke with Felipe Bravo, an academic at the Department of Computer Science (DCC) and director of the Data and Artificial Intelligence Initiative (IDIA) at the Faculty of Physical and Mathematical Sciences (FCFM) of Universidad de Chile, and a researcher at the Millennium Institute for Foundational Research on Data (IMFD) and the National Center for Artificial Intelligence (CENIA).
The conversation addressed the evolution of machine learning, the training of advanced talent in the region, and the conditions needed to build technological sovereignty from Latin America. Much of the dialogue, however, focused on a specific development: dashAI, the open-source platform that allows users to train and evaluate artificial intelligence models without writing code.
“The idea started before ChatGPT and consisted of having open-source software with an intuitive graphical interface, that is, no-code, without programming, that would allow users to carry out the entire life cycle of machine learning models. Loading data, exploring it, cleaning it, training classification models, regression models, image classification models, predicting customer churn, and so on,” Felipe explained as he recalled the origins of the project.
“That began because many projects were coming to the Millennium Institute for Foundational Research on Data (IMFD) and the National Center for Artificial Intelligence (CENIA), where we realized that people wanted machine learning models. But more than a model, they wanted to be trained to build their own models, because sometimes they had data (…) and for these people, navigating between PyTorch, scikit-learn, and Pandas was a very high training cost. So that is when we set out to build this software,” he added.
Weka, the reference point for dashAI
While pursuing his PhD at the University of Waikato in New Zealand, Felipe Bravo learned how Weka (Waikato Environment for Knowledge Analysis) worked, the first large-scale free and open-source machine learning software dedicated to machine learning and data mining.
“I understood very well the architecture of software of that complexity: that it had to be extensible, that it could be integrated with certain things, how to abstract certain things,” the researcher said about Weka. That tool, he explained, became outdated relative to current technologies, which is how the idea arose of rethinking the model with a web interface that would be extensible and compatible with contemporary libraries.
To work on usability and user validation, Claudia López joined the project in 2022. She is a specialist in human-computer interaction and also a researcher at the Department of Computer Science and the Data and Artificial Intelligence Initiative at the Faculty of Physical and Mathematical Sciences of Universidad de Chile, as well as a CENIA researcher.
dashAI: A local environment for sensitive data
dashAI is installed as a desktop application on Windows, macOS, and Linux, and it organizes work into modules. It allows users to load and explore data —common structured and multimedia files— and generate visualizations; train and compare classification, regression, or image recognition models; and access open-source generative models running on the user’s own machine. “The idea is to bring together different AI paradigms, classic machine learning alongside generative AI, within a single environment,” Bravo summarized.
That local architecture, he noted in the interview, is one of the reasons behind the interest the tool has sparked among organizations that work with sensitive information: because it does not depend on cloud services, data never leaves the device. Moreover, since it is open-source software, any institution can deploy its own version and adapt it to its needs.
More than 20 thesis students and a development method
The development of dashAI faced an initial funding challenge. “People did not understand what value an open, free software could have,” Bravo recalled when looking back on the project’s beginnings. The answer was to design a method that would incorporate students’ thesis work: each semester, four or five new members joined the project and dedicated a year to it—the first semester to formulating the idea, the second to developing it—supported by engineers who validated their progress. Whatever worked was integrated into the platform; whatever did not remained a prototype.
Unlike thesis projects that typically end up as isolated applications, this method allowed work to accumulate within a single platform. To date, more than 20 students have graduated as part of the project. The initiative later secured funding from ANID’s Fund for the Promotion of Scientific and Technological Development (FONDEF), through the IDeA I+D instrument awarded in 2025, under the public goods category.
dashAI was publicly launched in 2026 and, according to its website, has accumulated around 31,500 downloads. The platform is available at dash-ai.com.