DB Databricks Healthcare United States

BAYADA Home Health Care

Data Lake / Data Warehouse Modernization · System Migration

BAYADA delivers clinical, assistive, and hospice services to tens of thousands of patients weekly, but its legacy landscape of multiple practice management systems, an on-prem ODS, Snowflake, and legacy cubes forced analysts to reconcile conflicting numbers across systems for hours before every weekly review, slowing decisions about clients, caregivers, and operations.

3 separate data platforms consolidated into one governed warehouse
65+ enterprise data sources unified into a single Databricks SQL environment
~30% better workload performance and cost efficiency compared to Snowflake

Solution

BAYADA is rebuilding its analytics estate as a Databricks Lakehouse organized around mastered entities — Client, Worker, Visit, Payor, Office, Candidate, Referrer — using a medallion architecture (Landing/Raw, Bronze, Silver, Enriched, Master Data, Gold) to consolidate practice management systems, an on-prem ODS, Snowflake, Workday, RiskConnect, and marketing and experience tools. Lakeflow Jobs and Asset Bundles orchestrate ingestion and CI/CD across the tiers, while Unity Catalog enforces ownership, access, lineage, and quality. Databricks SQL serves the curated Gold layer to finance, operations, and clinical analytics as the single source of truth, and Databricks Assistant plus an LLM-powered code converter accelerate Snowflake migration. Genie extends natural-language access on the same governed foundation, preparing BAYADA for AI agents in payroll, compliance, and clinical workflows.

Data flow

Source systems — practice management platforms, an on-prem ODS, Snowflake, Workday, RiskConnect, and marketing/experience tooling — feed a standardized ingestion layer into medallion tiers (Landing/Raw, Bronze, Silver, Enriched, Master Data, Gold) on the Databricks Lakehouse. Lakeflow Jobs and Asset Bundles orchestrate pipelines and CI/CD, while Unity Catalog enforces ownership, access, lineage, and quality. Databricks SQL serves the curated Gold layer to finance, operations, and clinical analytics, and Genie plus Sigma provide natural-language and self-service access on the same governed foundation.

Solution architecture

6 components · 4 layers
  1. Compute
    • Databricks Assistant LLM-powered code conversion that automates large portions of SQL and stored-procedure migration from Snowflake
  2. Serving
    • Databricks SQL Warehouse performance on top of the Lakehouse, serving finance, operations, and clinical analytics as the single source of truth
    • Databricks Genie Natural-language access to mastered Client, Worker, Visit, and Payor entities
  3. Orchestration
    • Lakeflow Jobs Pipeline orchestration across the medallion tiers from raw landing through gold
    • Databricks Asset Bundles CI/CD packaging for jobs, pipelines, and shared assets across the platform
  4. Governance
    • Unity Catalog Ownership, access, lineage, and quality governance across the unified Lakehouse

Architecture clues

  • Entity-driven, event-based ingestion around Client, Worker, Visit, Payor, Office, Candidate, Referrer
  • FHIR-aligned canonical schemas
  • LLM-powered code converter + Databricks Assistant for SQL migration
  • Lakeflow Jobs + Asset Bundles for orchestration and CI/CD
  • ML-based data mastering accelerator
  • Medallion architecture: Landing/Raw → Bronze → Silver → Enriched → Master Data → Gold
  • T-Validator for automated integration test coverage
  • Unity Catalog for governance, lineage, and access control

Evidence from the source

3 Separate data platforms consolidated into one governed warehouse on DBSQL
65+ Enterprise data sources unified into a single Databricks SQL environment
~30% Better workload performance and cost efficiency compared to Snowflake