Free · open · 17 chapters
The FDE Handbook
A practical guide to forward deployed engineering: how to find the real problem, design the lightest system that solves it, prove it works, and land it with a customer. About 3.5 hours of reading, in order or by chapter.
Original scenarios
Every chapter is built around situations you'll meet with real customers.
Diagrams you can redraw
One picture per idea, drawn so you can sketch it on a whiteboard.
Exercises with answers
Commit to your answer first, then compare it with a worked one.
Credit where it's due
Public frameworks are named and linked to their authors.
The job
What forward deployed engineering is, and the shape of an engagement.
Understanding the problem
Finding the real problem behind the request, and writing it down so others can act on it.
- 03Discovery: the first meetingsWho to talk to, what to ask in which order, and how to leave with facts instead of a pitch. 14 min
- 04Reframing the askThe technology a customer asks for is a clue, not a diagnosis. How to find the problem underneath, and say so. 12 min
- 05Writing it downProblem briefs, acceptance criteria a stranger can test, edge cases, and decisions with an owner and a date. 13 min
Designing the system
Choosing the lightest tool, then building workflows, tools, retrieval and state that hold up.
- 06The lightest tool that worksRules, classifiers, models, workflows, agents: climb the ladder only as far as the problem needs, and price every rung. 11 min
- 07Workflows before agentsMost requests for an agent are better served by a workflow you design. Five patterns, three lanes, and exits to a person built first. 15 min
- 08Giving models hands: tools and integrationsLetting a model act on real systems safely: tool loops, honest results, protected writes, identity and MCP. 14 min
- 09Retrieval you can trustProduction RAG is a governed information system. Ingestion, permissions and freshness decide quality long before the model does. 16 min
- 10State, memory and agent boundariesWhere conversation state lives once there are two servers, and where to draw the lines when one agent becomes several. 12 min
Proving it works
Evaluation, unit economics and the controls that let a customer say yes.
- 11Evals: from demo to decisionDemos are picked; evaluations are not. How to build tests the customer believes, and turn them into launch decisions. 16 min
- 12Cost, latency and sizingPrice one unit of work, compare it to what the customer pays today, and size the system on a napkin before you build it. 13 min
- 13Security and how much autonomy to grantPermissions before context, untrusted text never in charge, and autonomy granted per action by impact and reversibility. 14 min
Landing it
Rollout, handover, and what to do when it breaks.
Getting the job
How FDE interviews work, and how to build proof you can point at.
- 16Interviews and what senior sounds likeThe rounds, the reasoning loop that carries you through open-ended cases, and the difference between a long answer and a senior one. 14 min
- 17Building proofProjects that look like deployments, stories with numbers, and how to tell a real FDE role from a relabelled one. 11 min
This handbook is written from scratch for Fieldwork and is free to read. Where a chapter builds on someone else's public work, the author is named and linked.