AWS re:Invent 2025-From fragmented to unified platform: Rocket Companies data and AI journey-IND3306

Transforming from Fragmented to Unified Data Platform: Rocket Companies' Journey

Overcoming Data Fragmentation Challenges

• Every enterprise today is accumulating more data than ever before, faster than their teams can organize it
• This leads to data fragmentation, with multiple sources of truth, inconsistent metrics, and difficulty finding the right data
• Fragmentation impacts downstream systems and teams, hindering the ability to derive insights and build accurate AI models

Rocket Companies' Transformation Journey

• Rocket Companies, a leading mortgage provider, recognized the need for a strong data foundation to support their rapid innovation and growth
• They identified three key challenges in their data landscape:
  1. Siloed and duplicated data across teams
  2. Difficulty finding the right data for use cases
  3. Challenges in utilizing rich data sets at scale for analytics, ML, and AI

Building a Unified Data Platform

• Rocket Companies made a strategic decision to move everything to a unified data lake, built on open table formats, standardized ingestion, and shared governance
• This was more than an infrastructure choice - it was an operating model and speed choice to enable rapid innovation

Key Components of the Unified Data Platform

1. Ingestion: Bringing in data from various sources (real-time streams, APIs, file drops, SaaS connectors) into a raw, immutable format on S3
2. Processing: Transforming the raw data once, but serving it to many use cases through a three-zone data lake (raw, processed, conformed)
   • Leveraging AWS services like EMR, Glue, and Flink for scalable, event-driven, and declarative data transformations
3. Consumption: Providing a unified service and single version of truth for all data consumers (analysts, AI/ML, operational apps)
   • Enabling self-service access and eliminating the need for teams to maintain their own data silos

Operational Excellence through DevOps and Automation

• Rocket Companies built a DevOps-driven, self-healing platform to automate the provisioning and management of the data infrastructure
• Defined reusable "envelopes" and deployment patterns to enable teams to provision new data pipelines and ML models in minutes, not weeks
• Implemented blue-green deployments, automatic rollbacks, and monitoring to ensure reliability and agility

Business Impact and Results

• Rocket's unified data platform powered a range of enterprise-grade solutions, including:
  • 360-degree APIs providing a single view of customers, mortgages, and transactions
  • Self-service, AI-powered "agentive" applications for business users
• Key outcomes:
  • 40,000 servicing leads integrated into the platform within 9 days
  • Reduced mortgage application to closing time from weeks to 3 days
  • 9-point increase in banker follow-ups and 10% lift in conversion rates
  • 20% increase in refinance pipeline during a market surge
  • 3x improvement in recapture rate compared to industry

Lessons Learned and Recommendations

• Adopt a three-layer data platform architecture: ingestion, processing, and consumption
• Leverage open data formats and AWS services for scalability, cost-efficiency, and operational simplicity
• Invest in DevOps and automation to enable self-service, reliable, and agile data infrastructure
• Focus on data curation and creating governed, trusted data products to accelerate downstream use cases
• Empower business users with self-service, AI-powered applications built on the unified data foundation