Forward Deployed Engineer Roadmap 2026
Work at the intersection of customer, product, engineering and AI. Master discovery, solution architecture, rapid prototyping, customer POCs, productionisation and enterprise AI deployment.
Visual Roadmap
11 stages · click any stage to jump
Software Engineering
The engineering craft customers assume y…
AI Engineering
The LLM, RAG and agent skills you will d…
Technical Discovery
Find the real problem, score use cases a…
Solution Architecture
Design under real constraints: permissio…
Rapid Prototyping
Prototypes that prove value quickly with…
Customer POC
Scope a pilot that cannot drift, and mea…
Production Deployment
Ship to real users inside the customer's…
Enterprise AI
Security reviews, executive communicatio…
Agent Deployment & Identity
Always-on agents with their own identiti…
Operating & Evaluating Agents
Evaluations, audit logs, cost controls a…
FDE Capstone
Run an end-to-end customer engagement on…
Stage by stage
0/21 chapters read · 0%1Software Engineering1 week0/3 readThe engineering craft customers assume you have.
You will be able to
- Build and deploy a typed API with a database
- Read an unfamiliar codebase quickly
- Integrate with an existing system through its API or database
Ready to move on when: You can stand up a small service against a database you did not design, in a day.
2AI Engineering2 weeks0/4 readThe LLM, RAG and agent skills you will deploy for customers.
You will be able to
- Build a RAG assistant with citations
- Give a model tools, and keep it safe
- Explain model behaviour and failure modes to non-specialists
Ready to move on when: You can build a cited question-answering prototype over a customer's documents in two days.
3Technical Discovery2 weeks0/3 readFind the real problem, score use cases and check the data.
You will be able to
- Run a discovery session that uncovers the real workflow and constraints
- Score and rank use cases on value, feasibility and risk
- Assess data readiness before promising anything
Ready to move on when: From one discovery session you can produce a ranked use-case list, with the data risks named for each.
4Solution Architecture2 weeks0/1 readDesign under real constraints: permissions, data residency, cost.
You will be able to
- Design for document permissions, data residency and on-premises limits
- Estimate cost and latency for the customer's volume
- Write an architecture decision record the customer signs off
Full series: FDE Engineering FundamentalsPractise
- Interview questionDesign a document assistant for 20,000 employees where people must only see answers from documents they are allowed to access.
- Interview questionA defence customer requires everything on-premises with no internet. How does your architecture change?
- Interview questionThe customer's CFO asks what the AI system will cost per month at full rollout. How do you estimate it?
Ready to move on when: You can defend an architecture, with its cost estimate, to the customer's security and finance teams.
5Rapid Prototyping2 weeks0/1 readPrototypes that prove value quickly without becoming throwaway work.
You will be able to
- Build a credible prototype on the customer's real or representative data
- Choose what to fake and what must be real
- Plan the path from prototype to production from the start
Ready to move on when: You can demo a working prototype on representative data within a week, with its production gaps listed.
6Customer POC2 weeks0/2 readScope a pilot that cannot drift, and measure it with the customer's data.
You will be able to
- Write a POC scope with a definition of done the customer signs
- Build an evaluation set with the customer's subject experts
- Handle scope creep and unrealistic accuracy targets
Full series: FDE Engineering FundamentalsPractise
- Interview questionMidway through a POC, the customer keeps adding "small" requests. How do you manage it?
- Interview questionHow do you build an evaluation set with a customer who has never done it?
- Interview questionThe customer says the system must be 100% accurate or it is useless. How do you handle that?
- Debug challengeFDE: 50,000 Document Assistant
Ready to move on when: Your POC has a signed scope, an evaluation set built with the customer and a pass mark agreed in advance.
7Production Deployment2 weeks0/2 readShip to real users inside the customer's environment.
You will be able to
- Move a POC to production in the customer's infrastructure
- Integrate with legacy systems
- Set up monitoring the customer's team can operate
Ready to move on when: The customer's own team can monitor the system and respond to an alert without you.
8Enterprise AI2 weeks0/2 readSecurity reviews, executive communication, handover and saying no.
You will be able to
- Pass a security review with a regulated customer
- Explain risks such as hallucination to executives without jargon
- Hand over cleanly, and say no to the wrong use cases
Read
Full series: FDE Engineering FundamentalsPractise
- Interview questionThree weeks in, the customer's security team says the project cannot send any data to an external model. What do you do?
- Interview questionAn executive asks: "Why does the AI make things up, and how do I know it won't do it here?" Answer as you would to them.
- Interview questionTell me how you would tell a customer that their favourite AI use case should not be built.
- Simulator scenarioLegal Research Assistant
Ready to move on when: You can take a project through a security review and an executive readout, and hand it over with documentation.
9Agent Deployment & Identity1 week0/1 readAlways-on agents with their own identities, scoped credentials and approval rules.
You will be able to
- Give each agent its own identity and least-privilege credentials
- Write an approval policy the customer's security team accepts
- Plan an agent pilot with clear exit criteria
Ready to move on when: You can present an agent's identity, permissions and approval policy to a customer's security team, and get a yes.
10Operating & Evaluating Agents1 week0/2 readEvaluations, audit logs, cost controls and incident response for agents in production.
You will be able to
- Monitor agent quality in production, and catch regressions
- Run an incident review for an agent that did the wrong thing
- Report agent value and cost to the customer
Full series: LLM EvaluationPractise
Ready to move on when: You can show the customer weekly agent quality, cost and incidents, and what changed as a result.
11FDE Capstone1 weekRun an end-to-end customer engagement on a realistic scenario.
You will be able to
- Run discovery to handover as one coherent engagement
- Produce artefacts a real customer would accept
- Reflect honestly on what you would do differently
Ready to move on when: Your engagement pack meets the brief below, and would survive a customer's review.
Capstone: An end-to-end customer engagement
Pick a scenario from the FDE simulator and run it as a real engagement: discovery, use-case selection, a scoped POC, a working prototype on synthetic data, a security review and an executive readout with handover.
What to deliver
- Discovery notes and a ranked use-case scorecard
- A POC scope with a definition of done and an agreed pass mark
- A working prototype on synthetic or public data, with its evaluation results
- Written answers to the ten questions a security team will ask
- A 10-minute executive readout and a handover document
What strong work looks like
- Problem framing
- The chosen use case clearly follows from discovery and scoring, not from the technology.
- Scope discipline
- The POC has a definition of done and a pass mark agreed before building.
- Technical delivery
- The prototype works on representative data, and its evaluation is honest about gaps.
- Trust and risk
- Security answers are specific, and risks have owners.
- Executive communication
- The readout leads with the decision and the evidence, with no jargon.
Start from: Insurance Claims Intake Agent (simulator scenario) · Your POC impressed everyone. List what has to change before it can go to production. (interview question)
What you'll be able to do
Series in this roadmap
Forward Deployed Engineer roadmap FAQ
How long does it take to become a Forward Deployed Engineer?
About 18 weeks of focused study on this roadmap, which has 11 stages. Engineers who deploy AI directly to customers — discovery, prototyping, integration and productionisation.
What does the Forward Deployed Engineer roadmap cover?
In order: Software Engineering, AI Engineering, Technical Discovery, Solution Architecture, Rapid Prototyping, Customer POC, Production Deployment, Enterprise AI, Agent Deployment & Identity, Operating & Evaluating Agents, FDE Capstone.
What will I be able to do after the Forward Deployed Engineer roadmap?
Run effective technical discovery with enterprise customers. Design solution architecture under real constraints. Ship customer-specific POCs rapidly. Productionise bespoke implementations. Communicate technical solutions to stakeholders.
How should I split my time on the Forward Deployed Engineer roadmap?
At 8 to 10 hours a week: Software Engineering 1 week, AI Engineering 2 weeks, Technical Discovery 2 weeks, Solution Architecture 2 weeks, Rapid Prototyping 2 weeks, Customer POC 2 weeks, Production Deployment 2 weeks, Enterprise AI 2 weeks, Agent Deployment & Identity 1 week, Operating & Evaluating Agents 1 week, FDE Capstone 1 week. Move on when you meet each stage's "ready to move on" checkpoint, not when the time is up.
What should I build to prove I am ready for Forward Deployed Engineer roles?
An end-to-end customer engagement: Pick a scenario from the FDE simulator and run it as a real engagement: discovery, use-case selection, a scoped POC, a working prototype on synthetic data, a security review and an executive readout with handover. Deliver 5 things, including discovery notes and a ranked use-case scorecard and a POC scope with a definition of done and an agreed pass mark.
Is the Forward Deployed Engineer roadmap free?
Yes. Every roadmap, series and progress tracker on Learn AI Engineering is free, and progress is saved in your browser without an account.