AI Agent
Revnue: AI Agent with PostgreSQL integration
"We partnered with Tech 42 to deliver a proof of concept for an AI-powered conversational agent built on AWS. The engagement focused on enabling natural language interaction with Revnue’s asset and contract lifecycle data using an agentic SQL approach, allowing users to query complex PostgreSQL-backed data through simple conversational prompts.
The solution leveraged AWS-native services including Amazon Bedrock, Bedrock Knowledge Bases with OpenSearch, LangGraph-based agent orchestration, and ECS-hosted APIs with observability through Langfuse. The architecture and codebase provided a clear, well-structured foundation that helped inform our internal technical decisions and next-phase planning.
Throughout the engagement, Tech 42 demonstrated strong technical execution, flexibility, and responsiveness. Communication was clear and effective, and the team was proactive in asking the right questions to understand our platform and constraints. Overall, the project successfully met its technical objectives and established a solid foundation for evaluating a future production implementation."
Sunny Sharma
Co-founder & CEO
Project summary
Revnue needed users to access complex asset and contract data without technical skill or developer support. Tech 42 built a generative AI agent on Amazon Bedrock with LangGraph orchestration and Bedrock Knowledge Bases backed by OpenSearch. Users query PostgreSQL-backed data through natural language prompts. The agent retrieves results using semi-parameterized SQL tools and surfaces answers in seconds. Bedrock guardrails filter content for compliance and safety. The system runs on ECS with auto-scaling and Langfuse observability to track agent performance. All infrastructure is defined in Terraform for reproducibility. The work replaced manual data access with a conversational interface that lets non-technical users answer complex questions independently.
