AI-Native Platform Engineering

The Client

Cigna Texas health insurance company is one of the major providers of health insurance to residents of Texas, offering both HMO and PPO plans, and competes with other leading health insurance companies licensed in Texas

Key Outcomes

Developed AI Engine for analysis of U.S. health insurance policies for PPO and HMO.
Developed a well-architected AI solution with NLP, OCR, and ML capabilities with a data ingestion pipeline and SQL database.

Business Challenges

To stay competitive, the client wanted to develop healthcare IT services that would help them compare their pricing and benefits with other competitors. However, they have many policy documents in PDF format, and manually extracting the relevant information could have been time-consuming and error-prone. The company needed a solution to automatically extract named entities and other relevant information from the PDF documents to facilitate comparison with competitors.

Business Impacts / Key Results Achieved

Cigna successfully achieved its goals by implementing healthcare IT services that leveraged AI-native analysis of U.S. health insurance policies, specifically for PPOs and HMOs. The client was able to utilize an ML engine with over 80% accuracy in evaluating policy coverage and costs and established a well-architected AI/ML services pipeline ensuring quality, security, operating-cost efficiency, and monitoring. Additionally, they generated clear financial reports on the UI console for management review.

Strategy and Solutions

Zymr AI/ML services experts and healthcare IT services experts collaborated to develop an AI-native application with complete lifecycle management to AI-ML and MLOps to deliver significant value for our clients. The solution was deployed on Azure and built an AI solution with NLP, OCR, and related ML capabilities to extract competitive analysis of benefits and coverage. Our data science team was engaged with the client to understand the business objectives thoroughly and developed the ML model in consultation with the business stakeholders of our client. Our MLOps team architected a data ingestion pipeline, storing raw, feature-engineered, and golden data set in a SQL database. THE NER ML model was developed iteratively using BERT and spaCy. Several tools and technologies were used, such as Tessaract, AWS extract for OCR, neural nets, and fast text for data classification. Kafka played a crucial role in the continuous ingestion and distribution of health insurance documents from various sources. These documents are subsequently processed by the AI engine developed by our AI/ML services experts. Initially, the ingested bronze documents undergo document classification, followed by preprocessing stages involving OCR, by tesseract, and AWS extract. Text classification techniques such as BERT are applied, along with named entity recognition (NER) using Spacy or custom NER. Post-processing involves JSON conversion and storage. The user interface (UI) is developed using Node and React, while MongoDB stores the Bronze, Silver, and Gold data sets.

Request A Copy
Zymr - Case Study

Latest Case Studies

With Zymr you can