How BlackRock Builds Custom Knowledge Apps at Scale
summary
BlackRock describes a Kubernetes-native AI framework with a sandbox and app factory that compresses complex document-extraction app delivery from about 3–8 months to a couple of days, handling thousands- to 10,000-page inputs and burstable GPU vs CPU inference clusters for scale. It uses template-driven extraction fields, validation rules, multiple LLM strategies and vendors, low-code transform/executor pipelines, CI/CD-style deployment, human-in-the-loop checks for compliance, and reports reduced operational overhead and faster generation of investment signals.
AI-written summary · about
chapters (8)
- 0:00Start
- 0:30Introduction to BlackRock's AI Initiatives
- 1:31Classifying AI Applications
- 2:22Use Case: New Issue Operations
- 3:59Challenges with Scaling AI Knowledge Apps
- 7:02Architecture of BlackRock's AI Framework
- 8:32Demonstration of the Sandbox
- 15:52Key Takeaways from the Discussion
related talks
34mKeynote: Building for AI at Scale: Infra, Integrity, & Innovation
18mKeynote: Adapting Kubernetes for Machine Learning Workflows
31mLamRAG: From Data to Constructive Insights using Amazon Bedrock
27mHarnessing Kubernetes for Scalable AI/ML Workloads
11mK8s-Native AI Microservices: Scaling in Regulated Industries
14mFrom 18 Months to 5 Days with Agentic AI on
26mLeveraging Internal Knowledge: Building AiKA at Spotify
31mAccelerate AI adoption: automated cloud efficiency