AI Platform Engineer

Our platforms turn messy source documents into numbers an investment committee or an accountant can defend. That only works if the AI layer is engineered like production software: measured, versioned, and traceable. You own that layer.

You will design extraction and reconciliation pipelines, build the evaluation harnesses that decide whether a model change ships, and integrate the results into the .NET and React platforms our clients run every day. You work directly with the founder and the senior team; what you build goes to production with your name on it.

• Document AI pipelines: extraction, validation, and reconciliation flows that turn source documents into structured, verified data.
• Traceability: every figure the platform produces links back to the document line it came from. You design and defend that guarantee.
• Evaluation: regression suites and accuracy benchmarks that gate every model or prompt change before it reaches production.
• Orchestration: model selection, prompt architecture, and cost and latency budgets across providers.
• Integration: AI services exposed cleanly to .NET Core back ends and React front ends.
• Operation: monitoring for drift and failure modes in live systems, and triaging what the monitors catch.

• LLM engineering: production experience with frontier model APIs (Anthropic, OpenAI, or Azure OpenAI), structured output, and retrieval.
• Languages: strong Python or C#; comfort reading and writing both.
• Data: SQL Server or PostgreSQL; designing schemas for verified, source-linked data.
• Cloud: Azure services for hosting, queues, and storage.
• Discipline: Git, code review, and testing habits that survive a regulated client’s audit.

• 5+ years of software engineering experience, including at least one AI or LLM-powered feature shipped to real users.
• Evidence you think in evals: you can explain how you knew your last AI feature actually worked.
• A portfolio of shipped systems counts more here than credentials.

Megsoft is a small senior engineering firm, founded in 2012 by a U.S. military veteran, that designs, builds, and operates production platforms: for Pepsi Co field teams, the NYPD, federal agencies, and the AI platform X CRE. Remote-first with headquarters in Odenton, Maryland.