Funded project, 2024 PredicTCR web service

The group want to provide limited access to their machine learning model which predicts tumor-reactive T cells and TCRs from scRNA-seq data, to allow other researchers to test the model without sharing the proprietary model weights.

For this purpose we created a website where users can register and submit sequences to be analyzed, as well as a distributed runner architecture for the model inference.

• Users can sign up for an account, log in, upload sequences, download their results
• Distributed runner system that provides docker images to do the model inference and upload the results
• Admin interface where user permissions, quotas, website settings and content can be updated
• Tech stack: vue.js frontend, flask backend, REST API, JWT authentication, docker
• Deployed as a set of docker services for portability
• Production docker containers are automatically re-built in Continuous Integration when code is updated