Roles
Hire AI Engineers on a Monthly Subscription
Hire AI engineers on a subscription to get agents, RAG pipelines, and MCP servers built and shipped to production, not just prototyped in a demo. Our AI engineers have built a live AI agent platform and a consumer AI chat product, and work alongside your team from 40 hours a month.
When you need AI engineers
Your board or customers are asking for AI features and nobody on your team has shipped an LLM product before.
You need a RAG system that answers questions from your own documents and data, not a thin wrapper around a chat API.
You want AI agents connected to your product's real data and actions through MCP servers, not a demo that only works in isolation.
You're already live with an AI feature and need someone to own quality: evals, cost control, and what happens when the model gets it wrong.
What they deliver
- AI agents wired into your product's real workflows, not standalone demos
- RAG pipelines over your own documents and data, with PostgreSQL or vector storage
- MCP servers that expose your product's data and actions to AI agents safely
- Voice agents for support or booking, built with n8n or custom code
- Evals and quality testing, so you know when a prompt or model change helps or hurts
- Deliberate choices on model provider, cost, and data privacy, not defaults
Stack and tools
What 40–160 hours looks like
40 hrs / month
One AI feature shipped end to end, such as AI-generated summaries on every support ticket or a FAQ chatbot wired into your existing docs.
80 hrs / month
A RAG system over your product documentation plus the evals to monitor it, or an AI agent that completes a multi-step task like triaging and routing a support queue.
160 hrs / month
A full AI product surface: agents, RAG, MCP servers, and the evals and monitoring to run all of it with confidence.
Subscription vs. hiring vs. freelancer
AI engineers are expensive to hire and hard to evaluate if you don't already have one on staff. A subscription gives you production AI work without the hire, the interview process you can't fully judge, or a freelancer's single point of failure.
| Subscription | In-house hire | Freelancer | |
|---|---|---|---|
| Start time | 48 hours | Weeks to months to hire | Days to weeks |
| Cost structure | Hours used against a monthly plan | [TBD: typical AI engineer base salary, dated source] plus benefits and recruiting | Hourly, rate and availability vary widely |
| Code review and testing | PR review, automated tests, CI/CD on every change | Depends entirely on the hire and your existing process | Usually none unless you set it up yourself |
| What happens if it's not working out | Cancel the plan | A bad hire costs months and is expensive to unwind | End the contract, but you've likely lost continuity |
| Evals and ongoing quality | Built into the engagement | Depends on the hire's experience with AI specifically | Rarely included |
The baseline most founders compare against: one full-time US developer costs $120,000–$180,000 a year in salary alone, before benefits, payroll overhead, or the months it typically takes to find and hire someone. AI engineering talent is scarcer and commands a premium on top of that baseline — we don't have a dated, sourced figure specific to AI engineer salaries yet, so we're not publishing one until we do.
Why this role is hard to hire for right now
Most SaaS companies adding AI features don't have anyone in-house who has shipped one. The skill is new enough that resumes are a weak signal — you're often evaluating someone's ability to talk about AI, not their ability to ship it. A subscription sidesteps that: you see working pull requests before you've committed to a single hire, and the risk of a bad fit is a cancelled plan instead of a six-month mistake.
What “production” means here
A lot of AI work never leaves the demo stage: it works once, in front of the right audience, with no evals, no monitoring, and no plan for what happens when the model is confidently wrong. Our approach treats AI features like any other production code — reviewed, tested, and measured — because that's the only way they survive contact with real users and real data.
FAQ
Questions about AI engineers
Do you build with OpenAI, Claude, or open models?
Whichever fits the job. We've shipped with OpenAI, Claude, and self-hosted Ollama models, and we'll recommend a provider based on cost, latency, and data privacy needs rather than defaulting to one.
Can you build a RAG system over our own data?
Yes. We've built this for appliance repair documentation (Appliance GPT) and can do the same over your product docs, support history, or internal knowledge base.
What's an MCP server, and do we need one?
An MCP server exposes your product's data and actions to AI agents in a safe, structured way, so an agent can actually do things in your system instead of just talking about them. You need one if you want an AI agent to take real actions, not just answer questions.
How do you handle data privacy with AI models?
Which data goes to which model provider is a deliberate decision, not a default, and we document it. See our Security page for our full approach to data handling.
How do you test AI quality?
Through evals: a repeatable test set that scores outputs against what a good answer looks like, so you can tell whether a prompt or model change made things better or worse before it reaches users.
Can AI engineers work alongside our existing development team?
Yes. Most engagements mix AI hours with full-stack or product hours on the same plan, and we work in your repo with the same PR review process as the rest of your team.
What's a realistic timeline for shipping a first AI feature?
A focused feature like ticket summarization or a documentation chatbot typically fits inside a 40-hour month. Larger builds, like a full agent with RAG and evals, usually need 80 hours or more.