How Bloomberg Uses AI to Help Speed up Over-the-Counter Trading
summary
Bloomberg deploys a pipeline that detects negotiated security offers inside group chats and emails, using Kubernetes CRDs plus KServe and Kubeflow for sampling, training, A/B tests and production. Large pretrained LLMs teach smaller low-latency models via model distillation so inference meets a millisecond latency requirement. Annotation yields are small (thousands per asset per year) yet the system supports hundreds of models, encrypted access controls, and reproducible end-to-end experiments on multi‑GPU Kubernetes clusters.
AI-written summary · about
chapters (8)
- 0:00AI in OTC trading
- 1:03What OTC trading means
- 3:49OTC trader communication
- 6:57Security offer detection with NLP
- 10:54Kubernetes in ML lifecycle
- 16:49Building models with fewer annotations
- 21:02Why Kubernetes for AI
- 22:26Q&A on batch processing
related talks
31mScaling LLMs at Bloomberg Using Envoy AI Gateway: A Hybrid Cloud Approach
08mQ n A with Dr Arun Verma
16mOperationalizing AI in Finance: Real-Time Analytics
18mKeynote: Adapting Kubernetes for Machine Learning Workflows
42mThe Challenges Managing a Kubernetes-Based Machine Learning Infrastructure
40mPanel: Kubernetes and GPU Trends for AI in Financial Services
27mHarnessing Kubernetes for Scalable AI/ML Workloads
13mScaling AI Systems for High-Volume Transaction Processing