How Jefferies Optimized Front-Office Trading Operations with an AI Assistant on AWS

Investment bank Jefferies has built an AI assistant for front-office traders running on AWS. The solution uses the Strands Agents agentic framework, Amazon Bedrock with Anthropic Claude LLM, Amazon Bedrock Knowledge Bases for RAG, and MCP tools to connect to data sources: in-memory databases, SQL databases, and FIX messages. A trader enters a natural language query, the LLM generates SQL, executes it, and returns results with visualizations. The system includes authentication via Amazon EKS, monitoring through Amazon Bedrock Guardrails for PII filtering and access control. To reduce hallucination risk, a separate engine generates visualizations rather than the LLM. In-memory databases ensure low latency. Jefferies plans global deployment of the solution. The assistant has already shown measurable efficiency gains, reduced IT time for dashboard creation, and democratized data access.