Search 15k clinical trials using machine learning - medBERT/NLP demo

Try it out by searching above for matching trials... try 'cancer' or 'asthma' or 'oxygen saturation' ... and click the search button. Note that some accurate matches do not contain the keywords you searched for, because language inference is taking the place of keyword search.

The backend is a Python web server running on a Google Compute instance, using machine learning (medBERT, an implementation of Google's BERT transformer) to pre-compute vector embeddings for the text abstracts taken from a database of public clinical trials, and compare the current query against that vector space, returning a dot-product distance 'score'.

This frontend is a Next.JS app running on Vercel. Take a look at the code on github at https://github.com/zachblume/machine-learning-recommendation-engine-demo

Type something in the search box at the top to begin...