AI Health Care Startup
Leveraging data architecture and NLP methods to combine, map, and match text and identifier hospital data to public medical and surgical data in order to deliver analytics and applications features for a venture backed AI startups product.
Background:
A venture backed AI technology company that works with hospitals (device and reporting for operating rooms) needed specialized data science and data architecture development to aid in building out analytics and application features in their product.
Challenge:
The challenges were threefold.
To introduce a method to combine public medical and surgical data for usage within the application.
Leveraging various NLP methods in order to map messy text and identifier client-side data to various public data bases.
Build a user facing strategy for various data and intelligence products.
Approach:
The AAARL team worked with the client over four months in 2023. Their CTO scoped out the project and our team utilized a parallel agile methodology to complete the project. Data engineering work was done through SQL then applied to cloud compute. The workflow was automated end to end. The data science project was completed in Python relying on variety of methods and then evaluated based on the framework our team developed. Fuzzy matching, TFID, word dictionary, minhash, and LSH were experimented and tuned.
Results:
The data engineering pipeline was ingested into the broader application environment. The word processing pipeline was handed off to the new data science team that will be used for a variety of application and business intelligence features.
The client was able to leverage our team instead of hiring a dedicated team of product manager, business analyst, data engineer, and data scientist. Saving an approximate 200~300k on development cost.
