Phase 0
Recon
2 weeks · Fixed fee
Map the environment, the integration surface and the political reality. Publish kill criteria. We will tell the vendor when a deployment should not proceed.
What We Do
A standing corps of forward deployed engineers, rented by the vendor, embedded at the customer.
We are sold to the AI vendor. We deploy into their customer. We are measured on time to production — not hours billed, not seats filled.
- The unit is a Pod — one lead FDE + two engineers, sharing a common platform and a deployment-IP library that starts filling from our first engagements.
- Co-located in India on a shifted schedule. Every engagement gets a named single point of contact on the client side.
- We take delivery risk on fixed scope and gated milestones. That is the whole difference.
How we differ
Not a staffing agency
Marketplaces transfer hiring risk to the buyer and stop there. The buyer still owns whether it works. Grapnel owns delivery outcomes.
Not a consultancy
Consultancies deliver a recommendation and a deck. We deliver a running system in the customer's production environment.
Not a dev shop
Dev shops build what you specify. We are handed a product that already exists and asked to make it survive contact with a real company.
How We Work
Four phases. Each one gated, each one designed to produce assets we keep.
Phase 1
Beachhead
6–10 weeks · Fixed fee, milestone-gated
One workflow, all the way to production, with real users. Not a pilot.
The single most valuable thing we sell
Phase 2
Expansion
Monthly pod retainer
Additional workflows, additional business units. This is where the account compounds and where retention is decided.
Phase 3
Handback
Enablement or continued run
Runbooks, evals and internal training — or we keep running it. Either way the connectors and playbooks will come home with us.
How We Select
Three buckets. Every FDE clears all three before they touch a customer.
Most staffing pipelines filter on one axis. We filter on three, because a forward deployed engineer who is strong on code and weak in the room fails a customer just as fast as the reverse.
We built and vetted the bench before the first deployment, not after — most delivery shops staff reactively per-deal. We didn't.
Bucket 1
Software Engineering
The floor, not the differentiator
Production-grade code under real constraints: can they ship into a live system, under a customer's review process, without breaking what's already running. We test this against real integration problems, not algorithm puzzles.
Bucket 2
AI Systems Fluency
Where most pipelines fall short
Can they reason about model behavior, not just call an API. Eval design, failure-mode triage, prompt and context engineering, knowing when a deployment is stuck because of the model versus stuck because of the integration — this is where most engineers, and most staffing pipelines, fall short.
Bucket 3
Customer Skills
The bucket nobody else screens for
Can they sit in a room with a skeptical VP of Engineering, tell them a deployment shouldn't proceed, and keep the relationship intact. Can they translate a customer's ambiguous complaint into a scoped fix. This is the difference between an engineer and a forward deployed engineer.
Clear all three, and you're in the Pod rotation. Clear two, and you're not — no matter how strong the third.