Protecting people's health and well-being with AI
Milliman's MedInsight group, which provides actuarial insights to over 300 U.S. healthcare organizations, replaced a rigid on-premises 50-node Spark cluster with the Databricks Data + AI Platform to process rising volumes of healthcare claims data. Using Delta Lake, Auto Loader and MLflow, Milliman MedInsight achieved 15x to 35x faster data pipeline performance and 5x to 7x faster delivery of health insurance risk insights, and plans to build an LLM-powered AI assistant trained on healthcare claims data and domain documentation.
Overview
Milliman's MedInsight group, which provides actuarial insights to over 300 U.S. healthcare organizations, replaced a rigid on-premises 50-node Spark cluster with the Databricks Data + AI Platform to process rising volumes of healthcare claims data. Using Delta Lake, Auto Loader and MLflow, Milliman MedInsight achieved 15x to 35x faster data pipeline performance and 5x to 7x faster delivery of health insurance risk insights, and plans to build an LLM-powered AI assistant trained on healthcare claims data and domain documentation.
This entry has 15 published fields tied to exact passages in an immutable source capture.
Inspect the highlighted sourceThe challenge
Milliman MedInsight's outdated on-prem database systems became too complex to maintain and too rigid to scale as healthcare data grew in volume and complexity. A small team tried building a 50-node cluster on Apache Spark but found it overwhelmingly laborious to maintain, equating every hour of customer-impacting work with four hours needed to maintain the cluster, and lacked the flexibility and resources to handle the scale and timeliness healthcare customers needed.
The solution
Milliman MedInsight adopted the Databricks Data + AI Platform, using Delta Lake as an optimized storage layer with different data tiers from raw data to cleansed data for reporting and machine learning. Auto Loader incrementally and efficiently processes new data files, and the data science team manages ML processes directly in MLflow, with decoupled compute and storage. The team built the Data Science Portal (DSP), a data science center of excellence giving customers self-service access to a library of data and models while removing the complexity of provisioning clusters, preparing data and ensuring security.
Reported business value
Milliman MedInsight has seen 15x to 35x improved workload performance, resulting in a 5x to 7x increase in the delivery of health insurance risk insights to customers, supporting the 300+ healthcare companies it works with. The company plans to build an AI assistant powered by large language models, trained on its domain expertise, documentation and terabytes of healthcare claims data.
Sources
Open any source and check the claim yourself — that is the point of the register.
This record was researched and written with AI assistance, and its claims were checked against the sources above. (EU AI Act art. 50 transparency notice.)
Other healthcare entries in the register.
Less paperwork, more time for people: Azure OpenAI at Landsberg Hospital
Klinikum Landsberg am Lech, a German acute-care hospital, deployed an AI language solution built on Azure OpenAI with partner Pexon Consulting that transcribes doctor-patient consultations in real time via tablet, recognizes medical terms, structures the content, and transfers it directly into the hospital information system.
Novo Nordisk builds AI drug discovery platform on Azure with Microsoft Research
Novo Nordisk partnered with Microsoft Research to build an AI platform on Azure AI and data stacks spanning regulatory affairs, early research, drug discovery and trial design, using Azure OpenAI Service, Azure Cosmos DB and Azure Kubernetes Service, with Power BI and Power Apps for collaboration. The platform includes a copilot for researchers, shared reasoning-chain templates, and governance/auditing of how data and models are used. The teams published early results on predictive AI models for cardiovascular disease risk detection, including an algorithm that Novo Nordisk says predicts patients' cardiovascular risk better than the best clinical standards, drawing on more than 100 years of insulin research data.
Novo Nordisk· DenmarkAzure OpenAI Service · Azure Cosmos DB · Azure Kubernetes Service +2BAYADA Builds a Unified, AI-Ready Platform for Compassionate Care
BAYADA Home Health Care is consolidating three separate legacy data platforms (multiple practice management systems, an on-prem ODS, and Snowflake) and 65+ enterprise data sources into a single Databricks Lakehouse under its Data Modernization program, using medallion tiers, Lakeflow Jobs, Asset Bundles, and Unity Catalog governance. As part of the migration, BAYADA uses an LLM-powered code converter and Databricks Assistant to automate SQL and stored-procedure translation, and applies a machine learning-based data mastering accelerator to create golden Client, Payor, Candidate, and Referrer records. BAYADA reports roughly 30% better workload performance and cost efficiency versus its prior Snowflake environment, and is laying a governed foundation for AI agents supporting payroll validation, compliance monitoring, and operational insights, separately from its roadmap for AI-assisted chart review and risk prediction.
BAYADA Home Health CareDatabricks SQL · Lakeflow Jobs · Unity Catalog +1Heidi Health fine-tunes NVIDIA Parakeet/Nemotron ASR, cutting clinical transcription latency 75% and costs 64%
Heidi Health, which processes 2.4 million clinical consultations weekly across 110 languages, fine-tuned the open NVIDIA Parakeet V2 speech recognition model on 1,500 hours of curated clinical audio using the NVIDIA NeMo framework and 8x H100 GPUs, replacing its closed-source vendor ASR. The in-house Nemotron-based ASR stack cut end-to-end transcription latency from about 3.0 seconds to 0.7 seconds (75%+ improvement), reduced word error rate from 13% to 9.4%, and cut operating costs by 64% by eliminating per-minute API pricing, while giving Heidi full ownership of model weights for data sovereignty.
Was this helpful?
Your feedback helps us improve our use case database