Laboratorio de Inteligencia Artificial

Paula Feldman

Doctoral fellow, CONICET

Abi Oppenheim

Research assistant

Hugo Massaroli

Research assistant

Martín Sinnona

Research assistant

Melisa Pandolfi

Master's student, MiM+Analytics, UTDT

Miguel Fainstein

Research assistant - Computer Science student, UBA

Santiago Corley

Research assistant

Francisco Leterio

Research assistant


Interaction between humans and AI systems

When engaged in conversation, human beings exhibit an extraordinary level of coordination along various dimensions of speech. This line of research seeks to understand and computationally model the different forms of coordination. The ultimate objective is to incorporate this knowledge into spoken dialogue systems (for example, virtual assistants) and thus try to improve their naturalness and usability for increasingly diverse tasks and audiences.

Automatic analysis of natural language

The immense quantity of data generated by humanity grows day by day. A considerable part consists of texts and audios in natural language (Spanish, English, etc.) posted on social networks. This line of research seeks to improve automatic natural language processing techniques, aimed at extracting valuable information from large volumes of data, both at the linguistic level (words, phrases, meaning, etc.) and at the paralinguistic level (emotions, sarcasm, violence, etc.).

Assisted manipulation of 2D and 3D content

Digital information gives us unprecedented abilities to create and edit content. However, some tasks on complex data types, such as images or 3D models, are still difficult and often require skilled personnel. This line of research seeks to improve digital manipulation tools through AI models, with a focus on CAD tools and industrial product design.

Social network analysis

The interaction with different devices and services that collect data from their users allows the relationships between them to be studied in detail. This line of research seeks to make use of modern network science techniques to model human behaviors that allow answering questions from disciplines such as Economics, Sociology or Social Psychology, as well as improving the performance of predictive models by incorporating this dimension into them.

Disease detection and understanding

Medicine generates multimodal datasets to improve disease detection, analyze treatment effectiveness, and in turn understand normal development. Data dimensionality ranges from 1D to 4D: from a PCR test (positive/negative) to magnetic resonance images acquired in space (3D), and over time for functional images (4D), and angular resolution for diffusion images (4D). This line of research aims to analyze and understand multi-modal medical data obtained in groups of patients in order to develop new models, and thus, improve the accuracy and detection of diseases as well as the understanding treatment effectiveness and disease progression.

Alternative economies in web3

In Latin America, the majority of workers are informally employed. They are excluded from the traditional financial system due to the lack of information to assess their creditworthiness. As a consequence, they are unable to access loans for growing their business or other services such as insurance. In recent years web3 (blockchain) created an architecture for the alternative, anonymous and decentralized financial system, reaching the interest of the informal economy of Latin America where the local fiat currency is often unstable. This line of research is focused on analyzing web2 and web3 data to promote an inclusive alternative financial system.


Laboratorio de
Inteligencia Artificial

Escuela de Negocios - Universidad Torcuato Di Tella
Avenida Presidente Figueroa Alcorta 7350 (C1428BCW)
Ciudad Autónoma de Buenos Aires, Argentina
Director: Agustín Gravano


Versión en español