Services
AI and ML, Digital Transformation, Quality Assurance
Industries
Healthcare & Life Sciences
Product
AI-driven service within the Medical Coding Interface that automates the work usually done by specialists or Medical Coders. It assigns medical codes (CPT, ASA, ICD-10) to clinical notes, significantly reducing the need for manual coding. AutoCoder streamlines healthcare billing and facilitates more efficient insurance processing with minimal human intervention by analyzing patient information, diagnoses, and procedures.
Business needs
- Achieve precise code assignment across various code types considering different strategies of insurance companies.
- Create AI-driven solution that requires minimum to no training when new data is provided.
- Enable assignment of correct codes with only a few data samples provided.
- Implement multi-layer validations for accurate final predictions.
- Develop an adaptable system for future enhancements and new features.
- Establish an automatic report generation framework for specific case subsets.
- Collaborate with SMEs to incorporate their expertise into the coding system.
Suggested solutions from svitla
- Developed and deployed a fully modular and configurable multi-layer Python service, enabling dynamic integration of AI and rule-based systems with adjustable execution sequences to optimize performance and scalability.
- Conducted R&D on replacing GPT with a locally hosted, low-resource consumption large language model (LLM) for efficient knowledge extraction, aiming to minimize operational costs and optimize processing speed without compromising the accuracy of insights derived.
- Enhanced input document understanding by implementing advanced preprocessing techniques to distill essential data, significantly reducing volume for processing and compacting context, thereby optimizing the efficiency of subsequent AI analysis stages.
- Leveraged AWS to automate task processing, orchestrate UI/API client responses, and deploy various AI solutions.
- Refined the system’s architecture, eliminating redundant processes and minimizing the required input files for operational efficiency.
- Implemented caching strategies to improve GPT utilization and reduce costs.
- Enhanced the base service with customizable sub-components for ongoing maintenance.
- Introduced scripts for automated report generation.
Technologies
Development and Testing: Python, Asyncio, unittest, DDT
Machine Learning frameworks: Pytorch, FAISS, Numpy, Pandas, Networkx
AI tools and frameworks: Azure GPT, Sagemaker, HuggingFace Transformers, SpaCy, Ray
Cloud Services and Deployment: AWS, Azure, Docker, PostgreSQL, S3
Value delivered
- Achieved significant cost savings and optimized GPT service use.
- Enhanced the system's capacity and scalability, improving overall maintainability.
- Increased the accuracy of code predictions and the robustness of validations.
See more cases
Mobile Apps For Jewelry Retailer
Discover how we developed an e-commerce mobile apps for a leading Latin American jewelry company with over 30 years of market leadership. With newly updated iOS and Android apps, our client successfully differentiated itself from competitors, driving significant increase in sales and revenue.
Automation For Drone Data Analysis
Discover how we developed an automated information system for a global conglomerate, streamlining data analysis from drone-captured media through microservices architecture and AWS cloud services, resulting in enhanced efficiency, reduced manual effort, and improved data visualization for Data Analysts.
NLP Enhanced Recruitment Solution
Explore how we transformed recruitment for an AI technology company by implementing a sophisticated user communication system and enhancing interview processes with real-time natural language processing, significantly boosting efficiency and enhancing candidate experiences.