This is a remote position.
Senior AI Engineer - ( MCP & LLM Integration)
About the Role
We are looking for a high-performing Senior AI Engineer (including MCP) to lead the development and delivery of enterprise Model Context Protocol (MCP) integrations in Phase 1, and governed multi-model access, evaluation, and observability in Phase 2.
The ideal candidate will combine strong backend software engineering in Python with direct hands-on experience implementing the Model Context Protocol (MCP), LLM APIs (Anthropic Claude, OpenAI), tool-calling, and evaluation frameworks, ensuring robust, production-grade agentic workflows and reliable model interfaces across the engagement.
Programme Context & Phased Effort Allocation
The engagement is structured as a 22-week sequential programme divided into two delivery phases, assuming a five-day working week:
- Phase 1: MCP Gateway and Governance (Weeks 1–10 / 10 Weeks): Full-time allocation (1.0 FTE) developing reference MCP servers, API adapters, central registry onboarding checks, and validating Claude client tool-use workflows.
- Phase 2: Model Access through Microsoft Foundry (Weeks 11–22 / 12 Weeks): Full-time allocation (1.0 FTE) integrating Claude, OpenAI, and specialist models, implementing streaming token protocols, usage attribution, and model evaluation harnesses.
Key Responsibilities
- Reference MCP Servers & Adapters: Develop reference MCP servers and API adapters, including JSON-schema tool definitions, granular permissions, input validation, and graceful error handling.
- Registry Integration & Onboarding: Implement MCP registry integration, automated onboarding checks, request/response logging, and telemetry ingestion.
- Client Tool Connectivity: Connect the selected Claude client to approved MCP tools and thoroughly validate representative agentic workflows and tool-calling execution.
- Multi-Model Provider Integration: Integrate Claude, OpenAI, and agreed specialist model endpoints through the approved Azure APIM gateway and Microsoft Foundry.
- Streaming, Attribution & Evaluation: Implement and test provider interfaces, streaming token responses, tool use, usage attribution, prompt caching, and model evaluation suites.
- Automated Testing & Documentation: Automate integration tests, resolve defects, and document reusable integration patterns and standard operating procedures (SOPs).
Required Skills & Experience
- 6–10+ years of software engineering experience with senior proficiency in Python (FastAPI, Pydantic, asyncio) or agreed backend integration languages.
- Model Context Protocol (MCP) & Agentic Workflows: Hands-on experience implementing the Model Context Protocol (MCP) specification, JSON-RPC, tool-calling schemas, and client-server integrations.
- Enterprise LLM Integration: Extensive experience integrating Anthropic Claude and OpenAI APIs, handling authentication, token rate limits, prompt caching, and streaming protocols.
- Model Evaluation & Observability: Experience with automated model evaluation, benchmark datasets, prompt engineering, tracing (LangSmith / OpenTelemetry), and cost/usage attribution.
- API Design & Automated Testing: Strong background in RESTful APIs, JSON schema validation, automated testing (pytest), mock services, and CI/CD integration.
- Security & Data Governance: Understanding of zero-trust API security, payload sanitization, credential handling, and enterprise data privacy.
Preferred Qualifications
- Capital Markets / Financial AI Applications: Experience deploying AI applications, document processing, or market research copilots in financial institutions.
- Microsoft Foundry & Azure AI: Experience with Microsoft Foundry, Azure OpenAI Service endpoints, and Azure APIM policies.
- Agent Frameworks: Familiarity with LangChain, LangGraph, AutoGen, or Semantic Kernel frameworks.
- Open Source Contributions: Active contributions to open-source AI tooling, MCP server repositories, or developer SDKs.
Key Competencies
- Model Context Protocol (MCP) Architecture
- Advanced Python Engineering & Asyncio
- LLM API Integration (Anthropic Claude & OpenAI)
- Agentic Tool-Calling & Schema Validation
- Model Evaluation, Benchmarking & Prompt Optimization
- Automated Testing & Integration Test Suites
- Production Telemetry & Token Usage Attribution
- Developer Tooling & Pattern Documentation