AI Platform Engineer - Agentic Middleware

Type: Contract (Contract-to-Hire available)
Location: Remote (Worldwide)
Hours: Full-time preferred (40 hrs/week)

About Us

We are a financial dashboarding company serving the veterinary industry. We hold deep domain expertise, proprietary charts of accounts across multiple veterinary practice models, and validated GL-coded repositories of veterinary services and inventory.

We are building an AI-powered middleware platform that lets our customers query their financial data through natural language — replacing traditional dashboard interaction with an intelligent, conversational experience.

The Founder is a DVM with extensive veterinary industry knowledge. You will report directly to the Founder and be the first technical hire on this platform.

The Role

You will design and build our agentic AI middleware layer from the ground up. This system sits between our customers and their financial data, using large language models and the Model Context Protocol (MCP) to deliver scoped, secure, intelligent responses — without ever exposing proprietary data or customers' sensitive financial information.

This is not a research role. You will be building production infrastructure that handles real financial data for real veterinary practices.

What You'll Build

• MCP servers that expose our proprietary GL coding repository, charts of accounts, and customer financial data as structured, scoped tools
• An orchestration layer that routes customer queries through the appropriate MCP servers and LLM calls
• Multi-tenant auth and permissions so each customer only accesses their own data plus relevant reference data
• Audit logging, rate limiting, and data access controls
• Integration points with veterinary ERP, practice management, and accounting systems
• A conversational, API-first interface that replaces traditional dashboard interaction

Requirements

Must Have:

• 5+ years backend/platform engineering (Python and/or TypeScript)
• Experience building multi-tenant SaaS platforms with strict data isolation
• At least one shipped production LLM-integrated application (RAG, tool-use agents, conversational AI with real data access — not just wrappers around ChatGPT)
• Strong API design skills (REST, GraphQL, or protocol-level work)
• Database expertise: SQL, query optimization, data modeling
• Security-first mindset — you think about auth, permissions, and data leakage before writing features
• Ability to read a protocol spec and implement it

Strong Plus:

• Experience with MCP (Model Context Protocol) — building MCP servers or integrating MCP into applications
• Familiarity with LLM APIs (Anthropic Claude, OpenAI) including tool use / function calling
• Experience with financial data, accounting systems, or GL structures
• LLM orchestration frameworks: LangChain, LlamaIndex, or similar
• Infrastructure experience: Docker, cloud deployment, CI/CD
• Understanding of data privacy requirements (SOC 2, data residency)

Nice to Have:

• Background in veterinary, animal health, or veterinary practice management software
• Experience in developer tools or platform engineering
• Contributions to open-source MCP projects or LLM tooling

What We Offer

• Fully remote, flexible hours
• Opportunity to define the architecture of a new product category from the ground floor
• Direct impact — you are not maintaining legacy systems, you are building the future of how veterinary practices interact with their financial data

How to Apply

Along with your profile, please include a brief note covering:

1. An LLM-integrated product you shipped and what you learned
2. How you would approach building a secure, multi-tenant AI query layer over sensitive financial data
3. Your familiarity (if any) with MCP, and how quickly you could get productive with it

Interview Process

1. Initial conversation (30 min) — background, mutual fit
2. Technical discussion (60 min) — architecture design session: "Design an MCP server that exposes GL lookup to an AI agent with per-customer permissions"
3. Take-home or pair programming (2–3 hours) — build a small MCP server or LLM tool-use integration
4. Final conversation with Founder — vision alignment, role expectations