Gustavo Landfried

Experience

• 

  Antropocaos

  Feb 2008 - Mar 2016

  Data Science
• 

  Buenos Aires University

  Feb 2010 - Jul 2010

  Graduate teaching assistance

  Seminar: Artificial societies and ethnography.Department of Anthropological Sciences.

• 

  Ministerio de Desarrollo Social de la Nacion

  Sept 2012 - Jun 2013

  Evaluation of Public Policies
• 

  High Performance Computer Lab

  Jan 2014 - Jun 2015

  Data Engineer
• 

  Universidad Nacional de San Martín

  Jun 2015 - Mar 2016

  Coordinator between the areas of Public Opinion and Informatics

  Within the PASCAL Programme, the body of the National Audiovisual Audience Measurement System responsible for the Buenos Aires Metropolitan Area, I was in charge of the administration of the database and the automatic survey system, coordinating the tasks of the Public Opinion and Informatics staff (social scientists and computer technicians).

• 

  Facultad de Ciencias Exactas y Naturales, UBA

  Feb 2016 - Jul 2022

  Graduate teacher assistance in Computer Science

  Director of master thesis in computer science and teacher in:- Algorithms and data structures I (with C++)- Introduction to Computer Science (with Python)- Functional Programming (with Haskell)- Computational Social Science (with R)

• 

  Instituto de Ciencias de la Computación

  Jun 2016 - Jun 2022

  PhD in Computer Science

  Release of the first open version of the state-of-the-art skill estimator, TrueSkill Through Time (TTT), creating the first packages available so far in Julia, Python and R. Unlike the models commonly used in the video game industry and academia, TTT propagates historical information throughout the entire causal network, providing estimates with low uncertainty at any given time, enabling reliable initial skill estimates, and ensuring historical comparability. Analytical approximation methods and message-passing algorithms allow inference to be solved efficiently using any low-end computer, even in causal networks with millions of nodes and irregular structures. Show less

• 

  Laboratorios de Métodos Bayesianos

  Sept 2022 - now

  Bayesian data science

  Decision-making in health, sports, education, and gambling based on model evidence.- HEALTH: Evaluation of the performance of diagnostic tests for Chagas disease in Latin America, in collaboration with the national health reference centers of Argentina, Bolivia, Colombia, the Fiocruz Foundation Brazil, the International Organisation for the Diagnosis of Diseases (FIND), and the European Cooperation in Science and Technology research network "Novel tools for test evaluation and disease prevalence estimation".- SPORTS and EDUCATION: Estimation of skill in the video game industry, high-performance sports, and educational systems at all levels. Developer and mantainer of state-of-the-art libraries for learning analysis in the Python, Julia, and R programming language communities (TrueSkillThroughTime), allowing for skill estimation with low uncertainty across the entire time series and ensuring the comparability of estimates across time and space.- GAMBLING: Maximization of resource growth rate over time in betting games or investments through diversification, cooperation, specialization, and heterogeneity strategies. Specification and evaluation of alternative causal models, computation of optimal beliefs given available information and predictions made by the contribution of all hypotheses. Content and advertising recommendation models. Show less

• 

  FIND

  Aug 2023 - Aug 2024

  Principal Statistical Advisor for a Latin American project of Chagas disease.

  Principal Statistical Advisor for the evaluation of diagnostic test performance for Chagas disease in Latin America, in collaboration with national health reference centers in Argentina, Bolivia, Colombia, and the Fiocruz Foundation Brazil, organized by the International Organisation for the Diagnosis of Diseases (FIND).Context: In Latin America, millions of people are infected with Chagas disease, but only 10% are aware of their infection. This disease is treatable if detected early. However, despite the absence of a perfect test, the current protocol mandates the use of expensive "reference" tests, which must be performed in triplicate at national reference centers located in major cities. This bottleneck is the primary challenge in Chagas diagnosis in Latin America.Objective: To evaluate alternative protocols for on-site Chagas diagnosis based on rapid tests.Methodology: The main challenge of this project involves evaluating the performance of diagnostic tests for Chagas disease based solely on their own imperfect diagnoses as observables. Even without knowing the true patient state (due to the absence of a perfect reference test), we develop Bayesian models capable of estimating the true performance of the tests. To validate our analysis, we also assess the performance of alternative causal hypotheses. To achieve this, we employ specialized Monte Carlo methods not supported by standard programming languages, which enable us to estimate the prior prediction of a dataset, P(Data | Model), and subsequently compute the optimal belief distribution about the alternative models, P(Model | Data). Show less

• 

  Facultad de Ciencias Exactas y Naturales, UBA

  Aug 2023 - Jan 2025

  Researcher (75%) and teacher (25%) in Computer Science

  Research 75%: Bayesian methods for specifying and evaluating causal arguments for decision making.Efficient inference methods for probabilistic evaluation of causal arguments expressed in natural language by players in judicial, environmental and epidemiological processes, among others. Development of agile methodologies based on causal graphical networks, which not only serve as an intuitive language for any person, it also provide the mathematical specification upon which the performance of alternative arguments is optimally computed given the available evidence.Teaching 25%:- Algorithms and data structures II (with Java).- Bayesian Inference.Other duties:- Director of master’s thesis in Computer Science. Show less

• 

  Universidad Nacional de San Martín

  Aug 2024 - now

  Profesor

  I am a professor of the Data Science degree program. In particular, I developed the syllabus for the course on Bayesian Causal Inference, which I am in charge of. This course focuses on evaluating alternative causal arguments through the (approximation to) strict application of probability rules, the reasoning system in contexts of uncertainty. The main objective of the course is to review the methods developed in recent decades to:• Mathematically specify causal arguments expressed in natural language using intuitive graphical methods.• Determine how the causal structure influences the flow of inference among the variables in the model.• Identify the causal effect between variables in a causal model based on observational data (without interventions).• Design experiments that allow for the evaluation of alternative causal theories.• Select optimal decisions in action-perception cycles with a hidden (simulated) nature. Show less

• 

  MUTT DATA

  Jan 2025 - now

  Senior Data Scientist

  We provide services to multinational companies related to Causal Inference problems, among others.