One thesis
Copyright ©
Thesis Pte. Ltd. All Rights Reserved
OFFICE
GET IN TOUCH
SHOWCASE
AI-LLM Web app for clinical trial
01 // Overview
For our client from DUKE-NUS, we have developed an AI-powered search engine platform utilising large language models (LLMs) for clinical trials. This platform is designed to assist patients in finding clinical trials that best match their preferences and conditions based on their own words. This approach helps overcome the barrier of not knowing specific medical terminology, ensuring patients can discover trials that may benefit them the most.
The design involved
- UI/UX design
- LLM interference implementation
- LLM prompt design optimisation for performance and cost
- Web app development (Frontend/Backend)
- Deployment using AWS.
02 // Key Achievements
Within a short development period of 2 months, we completed the design and developed valuable tools for debugging the LLM output.
03 // Challenges and Solutions
One critical challenge presented in this project:
- The LLM was based on the ChatGPT API, but the initial prompt design made it challenging to integrate the output into a fixed format for search display. We had to swiftly comprehend the original prompt design and optimise it for web front-end display.
04 // Impact and applications
This new platform will simplify the complex world of clinical trials for everyday citizens, reducing barriers and benefiting patients. Additionally, it will decrease the time and effort medical personnel need to invest in finding suitable trials.