AgentGraph pairs enterprises with vetted AI engineers to design and ship agents, MCP servers, and agentic systems. Builders join the network and are matched to client engagements that fit their expertise and working style.
Forward Deployed Engineers are the builders we place closest to the customer. You work embedded in a client's workflows, with the teams whose day-to-day the system will actually change, and you own the hardest part of an agentic project: deciding what is worth building.
What you will do
- Work embedded in client workflows, alongside the people whose work the system will change, until you understand the process well enough to argue with it
- Define and qualify AI use cases — separating the ones that will hold up in production from the ones that only sound good in a meeting
- Gather requirements and translate ambiguous business problems into technical direction a delivery team can act on
- Team consultatively with client stakeholders, from the operators doing the work to the executives funding it, and keep both audiences honest about what agentic systems can and cannot do
- Shape architecture, sequencing, and delivery plans, then hand off cleanly to the engineers building the system — or carry it yourself where the scope is small enough
How this differs from Applied AI Engineer
This role is more strategic than hands-on coding. You need to be technical enough to be credible with an engineering team, to read a codebase, and to prototype when a prototype settles an argument faster than a document. But your leverage comes from definition and direction, not volume of code shipped. If you would rather spend most days in the codebase, the Applied AI Engineer role is the better fit — the two overlap heavily, and many builders in our network do both.
What we look for
- Real experience embedded with customers — consulting, forward deployed, solutions engineering, or founding-engineer work where you owned the problem as well as the code
- Fluency with the current agentic stack: LLM application patterns, agents, MCP servers, retrieval, and evaluation, with an informed opinion about where each stops working
- The judgment to say a use case is not worth building, and the credibility to make that stick
- Writing and facilitation skills strong enough that your requirements survive contact with an engineering team
How engagements work
You join the AgentGraph network as an independent contractor rather than an employee. We match you to client engagements based on your expertise, availability, and the shape of the work; scope, rate, and expected commitment are agreed before an engagement starts. Work is remote, and engagements range from short scoped projects to sustained embedded work with a single client.
Compensation
Rates are set per engagement, based on scope, depth of expertise, and expected commitment. We are transparent about the rate before you accept any engagement.