AI Agent
Document Processing
RAG & Embedding
Spiegel Financial: Intelligent Document Processing and AI Agent built on Amazon Bedrock
"Tech 42 turned a complex AWS/Bedrock build into a working solution our team could offer to our end users.
Working with Tech 42 was a smooth and professional experience from start to finish. Their depth of expertise in AWS, Bedrock, and applied AI helped us move from concept to a functioning implementation that fit our specific business requirements. They took the time to understand our objectives, communicated clearly throughout, and stayed committed to getting the details right — making the collaboration both productive and genuinely enjoyable."
Damon Spiegel
Founder
Project summary
Spiegel Financial needed to extract structured data from quarterly capital account statements and capital call notices, and give their team a way to query an existing database without writing SQL. Tech 42 built an AI agent on Amazon Bedrock using the AgentCore architecture. Two Bedrock Data Automation blueprints handle financial document processing and store extracted fields into the existing statements table. A Postgres Text-to-SQL MCP server hosted on ECS converts natural language questions into database queries. A Bedrock knowledge base with an OpenSearch vector database supports document ingestion and retrieval. IAM provides secure system-to-system authentication. AWS ADOT delivers observability across the system. The project moved Spiegel Financial from concept to a working implementation their team could put in front of end users.
