Role Overview
Builds and enhances AI-enabled features within AI Launchpad — model integration, prompt/workflow design and evaluation — and works with the AI Agent & Marketplace POD to keep Launchpad AI features compatible with the Agent Marketplace's
Model Context Protocol (MCP) and
Agent-to-Agent patterns as they mature.
Key Responsibilities
- Develop and enhance AI-enabled Launchpad features, model integrations and prompt/workflow configurations (Schedule A.4).
- Run AI evaluations and experiments; produce evaluation outputs, prototypes and technical recommendations for ADM.
- Build retrieval and data patterns that support AI feature accuracy, grounding and reliability.
- Coordinate with the Marketplace POD's Agentic AI Engineer so Launchpad AI features can interoperate with MCP-based tool/context patterns as the Marketplace scope matures (Schedule B.6, B.7).
- Support AI-related technical documentation and analytics for Launchpad features.
- Assist with testing, defect triage and stabilization of AI-enabled components.
Required Skills & Experience
- 4+ years experience building applied ML/LLM-powered features in production software.
- Strong Python skills; experience with LLM SDKs/frameworks (OpenAI/Anthropic APIs, LangChain, or equivalent).
- Practical experience with prompt engineering, retrieval-augmented generation (RAG) and model evaluation methodology.
- Working understanding of API-based service integration and data pipelines.
Preferred / Nice-to-Have
- Familiarity with Model Context Protocol (MCP) concepts, tool-calling and agent orchestration patterns.
- Experience with enterprise data-source connectors (SharePoint, Teams, Outlook, Microsoft Graph).
- Exposure to agent evaluation, safety testing or hallucination/grounding assessment.
Collaboration & Reporting
Works day-to-day with the Launchpad Tech Lead and Flex Specialist; cross-collaborates with the Marketplace POD's Senior Agentic AI Engineer on shared AI patterns, subject to each POD maintaining its own Backlog and Sprint Commitment (Section 2.3)
