Factory's Forward Deployed Engineering Model for Droid
Source
AI Engineer
Author
AI Engineer
Date
Why it matters
It gives a practitioner-level playbook for deploying coding agents inside large, security-constrained codebases — customer-owned harnesses that stay model-independent and can run air-gapped, plus a scoring model for whether a codebase is even ready for higher autonomy.
Useful if you're deciding how far to push agent autonomy before the workflow collapses.
Factory CEO Eno Reyes explains how forward deployed engineers embed with large customers to feed real-world signals into Droid, the company's AI coding agent, treating the whole pipeline as a 'software factory' where signals become plans, get validated, and ship as outcomes.
He describes a customer-owned, model-independent harness that can run air-gapped, an autonomy maturity model that scores how 'agent ready' a codebase is, and a city-planning analogy warning that pushing autonomy too far too fast produces a showcase that nobody can actually live in.
Transcript
Factory's forward deployed engineers sit at the tip of the product, embedded with the largest customers and piping a constant stream of real world signals back into how the agent, Droid, gets built. Eno Reyes frames the whole thing as a software factory: signals flow in from the outside, become plans, pass through validation stages, and come out as shipped outcomes, a loop most orgs run badly until they actually invest in the system around it. The non negotiable piece is a model independent harness the customer owns, so the traces and the data stay theirs and Droid can even run air gapped inside their environment.
The frontier is autonomy. Factory scores a codebase on how agent ready it is, whether it runs linters and type checkers and how much of the work an agent can finish and verify without a human, and the payoff shows up on jobs like migrating 30 to 50 million line equities systems at a bank. Push too hard, though, and you hit the trap Eno Reyes borrows from city planning: an exemplar city built too far ahead of its time becomes a theme park, not a place people live. So the forward deployed job is a balancing act, moving customers up the autonomy curve fast enough to matter but not so fast that nothing holds, which is why the role now leans as much on business judgment as on engineering.
Speaker info:
- https://x.com/EnoReyes
- https://www.linkedin.com/in/enoreyes/
- https://enoreyes.com/
Timestamps:
0:00 - Introduction: where forward deployed sits
2:08 - The role inside the customer's environment
3:49 - The tip of the spear of the product
5:03 - The software factory: signals in, outcomes out
6:57 - Why you own the agent harness
8:11 - Air gapping Droid inside the customer
10:08 - The autonomy maturity model
12:05 - Making a codebase agent ready
14:53 - Migrating 40 million line codebases
16:38 - The city of the future analogy
18:34 - Constrained autonomy and legal droid
20:04 - Redefining the forward deployed role
