Energy Trading Platform Eliminates API Complexity with Autonomous AI Agent on AWS Bedrock
At a glance
A cloud-native ETRM/CTRM platform managing $100 billion in daily trading volume faced a 30% year-over-year surge in support tickets as users struggled to navigate nearly 100 complex APIs to query products and execute trade orders.
NMD deployed an AI agent using AWS Bedrock with Claude that autonomously orchestrates product queries and trade execution through natural language, eliminating rigid API documentation requirements.
The 6-week POC moved to production, establishing the customer as an early agentic AI adopter and laying foundation for fully automated trade brokering.
Industry
Use Case
AI Agents & Intelligent Automation, Workflow Automation, Natural Language Processing, Knowledge Management
Solution implemented
- MCP-Based AI Agent - Autonomous tool interactions using Python SDK (chat, api_translator, api_response_interpreter)
- AWS Bedrock with Claude - Natural language understanding, reasoning, and automated API payload generation
- Autonomous API Selection - Intelligently identifies and executes correct API from 100+ platform options
- Multi-Step Workflow Orchestration - Handles complex trade execution sequences requiring multiple API calls
- Infrastructure-as-Code Deployment - Terraform-based deployment with runbooks, cost models, and expansion roadmap
The value equation
- 6-Week POC to Production - Rapid deployment from concept to production-ready solution
- 30% Support Reduction - Eliminated API complexity reducing support tickets and operational overhead
- Market Differentiation - Early agentic AI adoption ahead of legacy energy trading competitors
- Automated Trading Foundation - Infrastructure enabling Phase 2 fully autonomous trade brokering capabilities
- Future-Proof Architecture - MCP standards ensure compatibility with emerging AI tools and platforms
Company Snapshot
The customer is a cloud-native ETRM/CTRM platform streamlining trading workflows across power, renewables, hedge funds, oil & gas, metals, and crypto. Supporting 50+ commodities and 25,000+ products with $100 billion in daily volume, the platform delivers real-time P&L, VaR, and portfolio data as a SOC-certified multi-tenant SaaS.
Location
Global
Customer Situation
The customer’s nearly 100 APIs created friction as users struggled with complex documentation and command-line interfaces to execute trades and retrieve market data. Users needed to query products (e.g., commodity symbols, contracts) and request trades (buy/sell orders) on the platform, but the technical complexity created barriers. A 30% surge in support tickets from higher volumes and diverse user needs threatened scalability.
In energy trading, platforms typically require extensive technical training for API functionality, creating adoption barriers and support overhead. Manual trade setup and exception handling eroded margins, with intervention costs misaligned with the customer’s SaaS model. The customer needed an agentic AI solution for self-service product queries and trade execution while maintaining multi-tenant security in a regulated environment.
NMD Solution
NMD and customer decided that they needed an agentic AI solution for self-service product queries and trade execution while maintaining multi-tenant security in a regulated environment. NMD implemented an AI agent using Model Context Protocol (MCP) enabling autonomous natural language interaction for product queries and trade execution while maintaining enterprise-grade multi-tenant security.
Enabling the AI agent required an MCP server using AWS Bedrock with Claude for contextual understanding, multi-step reasoning, autonomous tool selection, and a Chainlit frontend for testing.
Solution Deployed
- MCP-based AI agent with Python SDK supporting autonomous tool interactions (chat, api_translator, api_response_interpreter)
- AWS Bedrock with Claude for natural language understanding, reasoning, and API payload generation
- Autonomous API selection intelligently identifying correct API from 100+ platform options
- Multi-step workflow orchestration for complex trade execution sequences
- Infrastructure-as-Code Terraform deployment with runbooks, cost models, and expansion roadmap
Within 6 weeks, the customer’s agentic AI solution validated users could query products (commodity symbols, contract specifications) and request trades (buy/sell orders) using natural language without learning API documentation. The AI agent autonomously recognized request types, orchestrated multi-step workflows, generated API payloads, executed APIs, and returned conversational responses.
The MCP architecture provides forward compatibility with emerging AI standards, minimizing future refactoring. The customer achieved market differentiation as an early agentic AI adopter, positioning for efficient scaling while reducing support overhead. The proven MCP layer establishes foundation for Phase 2 automated trade brokering, enabling fully autonomous AI-driven execution workflows to further reduce costs and improve margins.
Ready to Transform Your Finance Business?
See how New Math Data can transform your finance business with AWS-powered innovation.