Forward-Deployed AI Architect
You sit inside the client’s problem, not beside it. This role designs and builds the first working system in someone else’s environment, on their data, under their constraints, and stays long enough to see whether it survives contact with real use.
Location
Remote, contracted from India, US or EU hours
Structure
Employment or contractor engagement, confirmed per country
Reports to
Delivery Lead
Travel
Occasional, client-dependent
Status
Open, rolling applications. Posted
The problem you will help solve
Most enterprise AI work fails somewhere between the demonstration and the second month of production. The model was fine. The retrieval was untested, the access controls were an afterthought, the failure modes were never written down, and nobody owned the system after handover. This role exists to close that gap: to build inside the client’s cloud, prove the system under conditions resembling real use, and hand over something an internal team can actually operate.
What you will do
Understand before building
Sit with the people who do the work today. Map the decision, the data behind it, and the cost of getting it wrong, before anyone talks about a model.
Design the smallest useful system
Choose the architecture that tests the risky assumption fastest, retrieval, evaluation, orchestration, and the boring integration work in between.
Build it in their environment
Systems run where the client’s controls require: their cloud, their infrastructure, or an approved provider. You work with their security and platform teams rather than around them.
Evaluate honestly
Build the evaluation set before the demo. Report what the system cannot do as clearly as what it can.
Make responsibility visible
Human review points, escalation paths, logging, and limits documented before deployment, not after an incident.
Hand over properly
Documentation, runbooks, and enough shared context that the client’s team can change the system without calling us.
What we are looking for
Substantial production engineering experience, you have operated systems, not only prototyped them. Fluency in Python and the surrounding production stack. Real experience with retrieval systems, evaluation, and the failure modes of LLM applications. The judgement to work in an unfamiliar domain without pretending to be an expert in it, and the confidence to tell a client their problem does not need AI.
We care more about the quality of your judgement and your work than about polished self-presentation. Send evidence.
Helps, but not essential
What we will tell you honestly
You will spend meaningful time on integration, access reviews, and stakeholder meetings, the work that makes a system real rather than impressive. Client priorities shift. Some engagements will end with us recommending against building. If that sounds like a waste of your skills, this is the wrong role.
Compensation and participation
The range for this role, the employment or contracting structure, and any equity or profit participation are shared in the first conversation and documented before work begins. We will not describe a role as senior to attract stronger candidates while offering junior authority.
Read about ownership and participation →Apply
Apply for Forward-Deployed AI Architect
Send the work, not the pitch. A résumé plus one thing you built, maintained, evaluated, or fixed tells us more than a cover letter.