I'm a Sr. Architect and Responsible AI Champ at Microsoft, Industry Solutions Delivery, Healthcare and Life Science OU and have extensive experience, over 24 years in the industry. I've always tied technology to their use cases and what problems it can solve and the business outcomes. I am very technical, a background in computer engineering, computer science, programming and development but approaching solutions only from a technical perspective hasn't always panned out without looking at how it'll be used and who's using it.
Idea:
- I wanted to learn hands-on Azure OpenAI resources, model deployments of i.e. chat model (e.g. gpt-35-turbo-16k, gpt-4) with some kind of use case in mind.
- Since I'm in the Healthcare and Life Science industry, I was looking for use cases that could be helpful and ended up asking "Is there a way to search Clinical Trials data using ChatGPT?" Of course, there are other ways to search clinical trials via ClinicalTrials.gov and the dataset is public but wanted to combine OpenAI + AI Search + ChatGPT.
- So, I embarked on answering this question and learning along the way ended up deploying the following components:
- Azure OpenAI Service
- Blob Storage (Several ways to get data here: https://classic.clinicaltrials.gov/ct2/resources/download. I just went with a simple storage so I can house the XML files)
- Azure AI Search Index
- A Chat App UX (Luckily there is a sample app available I can use to start from https://github.com/microsoft/sample-app-aoai-chatGPT/). I've cloned that repo to start the clinical trials copilot repo.
- Azure CosmosDB to store chat history (this is also included in the sample AOAI chat app)
- PowerBI to view chat data from CosmosDB
Conceptual Architecture:
See the full details in my repo, Enjoy! https://github.com/dondinulos/clinical-trials-copilot/
Updated May 15, 2024
Version 2.0
dondinulos
Microsoft
Microsoft
