Most developers are told to specialise in a framework and wait for a promotion. Meanwhile, one role has quietly become the fastest-growing job in tech: the forward deployed engineer, or FDE. It rewards builders who can ship working software inside a customer's messy reality, not just inside their own codebase.
If you already write code and enjoy talking to the people who use it, this role is worth understanding now. Here is what it is, why demand is spiking, the stack to learn, and a 90-day path to get there using the iHateReading AI Agent Developer Roadmap.
What is a forward deployed engineer?
A forward deployed engineer is a software engineer who writes production code for a customer's problem rather than for their own company's general product. You embed with the customer, learn their workflow, and build or integrate something that works in their environment. Palantir coined the title in the mid-2000s by placing engineers directly with clients.
The simplest way to separate it from nearby roles: a product engineer builds one capability for many customers, and a solutions engineer owns the pre-sale demo and proof of concept. An FDE owns the outcome after the contract is signed. Titles are blurring, so judge a posting by its content. If most of the job is coding and owning production, it is an FDE role whatever the label says.
Why the demand is exploding
AI is the main driver. Companies can buy a model in minutes, but connecting it to legacy systems, messy data and real workflows is hard, and that gap needs people on the ground.
The numbers are large, though they vary by source. Indeed data reported by LeadDev shows FDE postings up more than 729% year over year in April 2026, with an average base salary near $172,000. Paraform's hiring analysis cites a median base around $174,000 and growth above 1,000% in an analysis of 1,000 roles. Treat these as US figures and as directional, since relabelled solutions-architect jobs inflate the counts.
Two honest caveats. Some roles demand heavy travel, and the same LeadDev piece notes an OpenAI spec asking for up to half of your time on the road. And salary in India or other markets will differ from these US numbers.
The tech stack to learn
You do not need every tool. Hiring guides keep converging on the same practical core: Python, SQL, APIs, a cloud platform, Docker, Git and modern AI tooling. Map that onto the layers below, each of which has a matching section in the roadmap.
- Foundations: Python for agents and data work, TypeScript for customer-facing apps, SQL for the data you will always need to query.
- LLM APIs: OpenAI, Anthropic and open models. Learn structured outputs and function calling first.
- Agent frameworks: compare options in our Python AI SDK breakdown and the roadmap's frameworks page.
- Tools and MCP: connecting agents to a customer's systems through function calling and the Model Context Protocol is the daily work.
- Retrieval: vector databases and RAG, so an agent can answer from the customer's own documents.
- Execution safety: agents that run code need isolation. Read why in AI agent sandbox SDKs and APIs.
- Reliability: the loop, memory and tracing around the model matter more than the model. See Agent Harnessing.
- Deployment and cost: Docker, one cloud (AWS or Azure), CI, and a habit of tracking spend per run.
The skills that actually get you hired
Code gets you the interview; judgment gets you the offer. Hiring checklists for FDEs emphasise end-to-end ownership without waiting for handoffs, plus the ability to translate a business problem into a technical plan and build trust with both engineers and executives.
Practise three habits. Scope vague problems into a small first deliverable. Explain trade-offs to a non-technical person without jargon. And always ask what happens when the integration fails at 2 a.m. Our field notes in Onboarding Old Industries show how this plays out: owners respond to "this saves your team four hours" far more than to feature lists.
Where frontend developers fit
If you came up through React, you have an underrated edge. Customers rarely want a notebook or an API; they want a screen their operations team can actually use. A fast, clear dashboard on top of an agent is often the difference between a demo and adoption.
So if you are frontend-leaning, keep that strength and add the backend and AI layers around it. Use the frontend roadmap for the interface side, and keep performance honest with bundle size discipline. Then learn enough Python and retrieval to own the whole path from data to screen.
A 90-day plan to become a forward deployed engineer
Common transition paths start from backend, full-stack or platform engineering, then add the AI stack and customer exposure. A realistic sequence:
- Days 1-30, learn the agent basics. Work through the first half of the roadmap: languages, LLM APIs, function calling, and one framework. Build a small tool-using agent.
- Days 31-60, add data and integrations. Add retrieval over real documents, connect two external APIs, and write tests for the failure cases. Deploy it somewhere public.
- Days 61-90, work with a real user. Find a small local business or a friend's team and automate one painful workflow. Document the problem, your scope, the result and what broke. This case study is your portfolio.
A guide on breaking into FDE roles makes the same point: something deployed that other people actually use beats a folder of tutorials. Rehearse with the roadmap's AI agent developer questions before interviews, and read about AI agent compliance because enterprise customers will ask.
Is it right for you?
The role suits people who like ambiguity, travel or frequent context-switching, and direct customer feedback. It suits you less if you want long stretches of solo, uninterrupted focus. Many FDEs also move later into solutions architecture, product or founding-engineer roles, so even if you do not stay, the skills compound.
Start building this week
The shortest route is to pick one workflow, build one agent, and put it in front of one user. Open the AI Agent Developer Roadmap, pick the next unchecked topic, and ship something small by Sunday. When you are ready to look at openings, browse the curated Jobs board and the monthly iHateReading Magazine for what is moving in the stack.