Matthew Le Email and Phone Number

Research engineer at Facebook AI Research (FAIR) working in the core machine learning group.COVID-19 ForecastingServed as the lead engineer for FAIR's COVID-19 forecasting initiative. Forecasts were generated on a daily basis and were made publicly available via the Humanitarian Data Exchange. Forecasts were included in a weekly report provided to the CDC.Sparse Multivariate Hawkes ProcessesBuilt scalable implementation of Sparse Multivariate Hawkes Processes (MHPs) that takes advantage of a common sparsity characteristic in many temporal diffusion proceses. Built a custom CUDA kernel callable from PyTorch to scale the likelihood computation. Published paper in the WebConf 2021.Low Resource Machine TranslationExplored techniques for improving the quality of neural machine translation in low resource settings. Submitted to, and won the 2019 Workshop on Asian Translation for the Burmese-English Track.Poincare EmbeddingsPrimary developer/maintainer of the open source poincare embeddings repository. Worked on scaling these methods to work on graphs with millions of nodes and billions of edges. Published paper in ACL 2019 on inferring concept hierarchies from unstructured text.

Using convolutional neural networks to automatically identify isolated communities in low-resource settings to aid community health organizations in healthcare delivery. Using AWS S3 for storing satellite imagery, PostgreSQL/PostGIS for managing spatial vector data, and AWS EC2 GPU instances for training models. Models are being developed using the PyTorch library.

• Worked on the Manticore project, a compiler for a parallel/concurrent dialect ofStandard ML.• Research focused on using first class continuations to reduce the overheadof software transactional memory.

• Designed and implemented algorithms for detecting and tracking ocean eddies