Jefferies builds AI trade assistant on AWS for real-time equities data queries

Jefferies, a global investment bank, has deployed an agentic AI trade assistant on AWS to revolutionize how its front-office equity traders interact with data. The solution, built with Strands Agents SDK, Amazon Bedrock, and Anthropic Claude, allows traders to pose natural language queries and receive real-time insights from millions of rows of trade data, FIX messages, and in-memory databases. The system uses Model Context Protocol (MCP) to securely connect to diverse data sources and employs Amazon Bedrock Knowledge Bases with Titan Embeddings for semantic retrieval. Conversation context is maintained across sessions, enabling traders to drill down on topics without needing to rephrase or re-query.

Key technical components include an authentication service on Amazon EKS, a bot service managing session context, and a query agent that translates natural language into SQL using Claude. The solution also incorporates Amazon Bedrock Guardrails for PII filtering and content moderation, plus row-level data entitlements to enforce access controls. By eliminating the need for traders to wait in IT queues or build custom dashboards, Jefferies has reduced the time to insight from days to seconds, closing the gap between available data and trading decisions. The assistant is embedded directly into Jefferies' existing Global Flow Monitor (GFM) interface.