Ricardo Dodds

Senior Ml Engineer @Memorial Sloan Kettering Cancer Center

NY, US

Currently working on the deployment of ImPartial, an active learning framework for assisted pathology slides segmentation. This project aims to allow pathology experts to label complete datasets based on the input of a few scribbles. We translated a research implementation of the model into a production ready API that ties together training, inference and sample selection strategy. Built the API on top of MONAI Label, becoming one of the first pathology use cases using this open-source NVIDIA platform.Recently developed the deployment pipelines for DeepLIIF, a huge multi-task deep-learning model composed by 9 networks, including both Resnets and Unets. Used Pulumi to define the cloud architecture, so that our rather small team could easily deploy new updates. The model is publicly available both through a web application https://deepliif.org and also through an API for programmatic use cases. Deployed the model on top a task queue for larger request and an auto- scaling cluster so that resources (GPUs) increase based on the traffic. This work was presented at the last edition of CVPR in New Orleans.

Sapienza Università di Roma

Doctor of Philosophy (PhD), Computer Science

Universidad de Chile

Bachelor of Applied Science (BASc), Ciencias naturales y exactas

Universidad de Chile

Engineer's degree, Ingeniería eléctrica y electrónica