Xavier Sumba
I am a Senior AI Reseach Developer at Heyday, acquired by Hootsuite, working in Conversational AI and IR ML. I did my masters at Concordia University and McGill University, where I focused on Approximate Bayesian Inference working with count data. I designed models using topic models and mixture models for NLP and healthcare. I've also worked with Semantic Web technologies (check out one of my projects). I became a member of the Apache Software Foundation, and I participated in Apache Marmotta. I did my undergrad in Computer Science in Ecuador at the Universidad de Cuenca.
My current research interests lie in the interaction of NLP, GRL, and Bayesian Inference. First, NLP offers a bridge so that computers can understand human language and carry the meaning we have collected during years of evolution. Graphs, on the other hand, provide a flexible framework to represent information heterogeneously. Finally, intelligence comes with the ability to understand knowledge. The Bayesian framework is a crucial component for decision-making and quantifying uncertainty. Indeed, other AI subfields (e.g., lifelong learning and meta-learning) can be seen through the Bayesian lens. Yet more powerful, it offers the ability to formulate causal questions. Additionally, since most advances in AI have been due to DL, it cannot go unnoticed. As a result, I am interested in treating these areas through a holistic approach and making these solutions accessible for solving real-life problems, particularly for social science and healthcare, to improve decision-making.
selected publications
- MLClustering count data with stochastic expectation propagationIn Asian Conference on Intelligent Information and Database Systems 2021
- DLBetween the interaction of graph neural networks and semantic webIn Proceedings of the 2019 NeurIPS Workshop on Graph Representation Learning 2019
services
Google Summer of Code | Port Apache Marmotta to Eclipse RDF4J 2017 |
---|---|
Volunteer | NeurIPS 2018 , ICML 2019 , EMNLP 2020 , ACL 2020 , NeurIPS 2020 , NeurIPS 2021 |
LatinX in AI Chair | ICML 2019 |
LatinX in AI Program Committee | ICML 2019 , NeurIPS 2019 , ICML 2020 |
Reviewer | LatinX in AI workshop at CVPR 2021 2020 , ML Reproducibility Challenge 2020 , ML4H: Machine Learning for Health 2021 , ICBINB@NeurIPS 2021 |