FirstClass Healthcare
FirstClass Healthcare is redefining healthcare delivery in correctional facilities through a unified platform spanning administrative, clinical, pharmacy, and finance operations. Within finance, a comprehensive claims management module streamlines end-to-end medical claims processing, with an AI-powered claims parsing tool at its core.
The story of FirstClass Healthcare
FirstClass Healthcare was founded with a mission to transform medical and mental healthcare in correctional facilities through compassionate, clinician-led, patient-centered care. Operating across medical, mental health, pharmacy, dental, and finance functions, FirstClass Healthcare addresses the disjointed, manual, and compliance-heavy environment of correctional healthcare.
Partnering with Carbonteq, they developed a unified digital platform to streamline operations, improve efficiency, and ensure compliance. Today, FirstClass Healthcare uses this platform to manage healthcare delivery across six county facilities in Georgia, delivering high-quality, coordinated care while optimizing operational performance.
Challenges faced by FirstClass Healthcare's finance department
Time-intensive billing and data entry
Processing medical claims required extensive manual data entry from complex documents, making billing slow, error-prone, and heavily reliant on human effort. Teams spent more time inputting data than managing and closing claims.
Disconnected systems and insurance workflows
Claims, billing records, and insurance details were managed across multiple systems with no unified workflow. Linking claims to insurance providers was cumbersome, leading to delays, reconciliation issues, and fragmented operations.
No visibility into the claim lifecycle
There was no centralized way to track claims across submitted, processed, approved, and paid stages. This limited visibility made it difficult to monitor paid claims, identify bottlenecks, and make informed financial decisions.
No intelligent validation or automation
Critical checks such as duplicate claim detection and accurate payable-amount calculations based on total charges and discount rates were handled manually, increasing the risk of errors, inconsistencies, and rework.
Scaling with growing volumes
As FirstClass Healthcare expanded across multiple county facilities, claim volume increased significantly. Existing manual processes struggled to keep up, slowing claim closures and limiting overall operational efficiency.
Carbonteq's tailored AI solution
Carbonteq designed and built an end-to-end AI-powered claims processing system focused on automating medical claim data extraction and improving operational efficiency. The solution goes beyond simple OCR by combining document classification, intelligent parsing, structured data extraction, model fine-tuning, and integrated workflows, enabling claims to be processed directly inside the platform with less manual effort.
- •
Hybrid AI claims parsing engine: The claims parsing engine uses a hybrid AI approach instead of relying on a single model for every document. Gemini 3.6 Flash handles complex and varied claim formats, while a smaller fine-tuned Gemma 4 model processes high-volume CMS-1500 claims, providing broad document coverage and a specialized path for the most common claim type.
- •
Dedicated claim classification service: A separate classification service identifies the uploaded document type before extraction begins. It recognizes CMS-1500, CMS-1450/UB-04, and vendor-invoice claims such as GDS (Global Diagnostic Services), which do not follow a fixed layout, allowing classification and extraction to scale, evolve, and be tested independently.
- •
Fine-tuned extraction model: Using previously processed CMS-1500 claims and stored AI responses, Carbonteq created a dedicated training dataset and fine-tuned a compact Gemma 4 E4B model. It learned claim structure, field locations, and the expected output format, returning accurate structured data with lower latency and operating cost than a larger general-purpose model.
- •
Deployment and smart routing: The fine-tuned CMS-1500 model is deployed on RunPod for live inference. When the specialized model is ready, CMS-1500 documents are routed to it for faster extraction. If it is warming up or unavailable, the system automatically routes the claim to Gemini so processing continues without blocking the user.
- •
Live inference and workflow integration: A live inference pipeline accepts claim PDFs and returns structured claim data through a dedicated extraction endpoint. The parsing engine is integrated directly into daily finance and operations workflows, eliminating manual data entry and accelerating claim processing.
- •
Human-in-the-loop review and AI-assisted labeling: A custom internal dashboard lets teams review claims individually, see the original PDF alongside extracted data, and efficiently create or edit ground-truth labels. Reviewers validate and correct AI predictions instead of labeling each document from scratch, improving medical billing accuracy while creating data for future benchmarking and fine-tuning.
- •
Format-specific extraction and latency optimization: Each claim type uses tailored extraction instructions and validation rules rather than a generic prompt. The system classifies each document first and routes it to the most suitable extraction path, improving reliability across formats while reducing unnecessary processing and latency.
- •
Evaluation, benchmarking, and continuous improvement: Predictions are stored alongside curated ground truth and scored with field-level metrics for consistent measurement of extraction quality and side-by-side comparison of model runs. AI inference, dataset management, human review, and evaluation together turn historical operational data into a continuously improving claims processing system.
FirstClass Healthcare & Carbonteq's partnership snapshot
The partnership between Carbonteq and FirstClass Healthcare marked a key step in modernizing its financial operations. Carbonteq designed and built a comprehensive claims management solution integrated into the platform. At its core, an AI-powered parsing engine processes medical claims across diverse formats, extracts and structures data, and matches it with existing records and insurance information. The result is a streamlined, intelligent workflow that enables efficient, scalable claims management across multiple facilities.
Results
Fully digitized, scalable operations
Spreadsheet-based processes were replaced with an AI-enabled digital workflow, allowing seamless scalability across multiple correctional facilities.
Significant reduction in manual effort
Medical claim data entry was largely automated, reducing dependency on manual input and allowing the finance team to focus on higher-value tasks rather than repetitive processing.
Faster, more accurate claim processing
AI-driven parsing and validation improved processing speed while minimizing errors, reducing rework and enabling quicker claim closures.
Enhanced visibility and control
A centralized, intelligent system provided clear processing and tracking across the claim lifecycle, including submitted, processed, and paid claims, improving financial oversight.
Carbonteq's AI-driven claims solution transformed FirstClass Healthcare's financial operations, turning a previously manual and time-intensive process into a streamlined digital workflow.
Work with us
Ready to build something that lasts?
Every engagement starts with a real conversation. No proposals, no pitch decks. Just an honest look at what you’re building and whether we’re the right team to build it.