What a forward deployed engineer does
The forward deployed engineer (FDE) is the engineer who works directly with a customer, inside their world, to turn a powerful technology into something that actually solves their problem.
What an FDE does
Most engineers build a product and hand it to customers. An FDE does the opposite: they go to the customer and build with them. In practice that means:
- Discovery: working out what the customer really needs, which is often different from what they asked for.
- Building: writing production code, integrations, data pipelines and AI workflows inside the customer's environment.
- Shipping: getting a working system into production, with the security, reliability and governance an enterprise demands.
- Feeding back: turning what they learn in the field into improvements to the core product.
Where the role came from
The title was popularized by Palantir, which sent engineers to sit with customers and make its software work on their hardest problems. Generative AI made the role far more widespread: models are powerful, but connecting them to an organization's data, systems and rules is hard, specific work that someone has to do on site.
How it differs from similar roles
| Role | Main focus | Writes production code? |
|---|---|---|
| Software engineer | Builds the core product | Yes |
| Solutions architect / sales engineer | Designs and demonstrates solutions before and during a sale | Sometimes, mostly prototypes |
| Consultant | Advises and delivers projects | Varies |
| Forward deployed engineer | Owns the customer's outcome from problem to production | Yes, in the customer's environment |
Because titles overlap, read the job description: if it says you'll own delivery and write code that customers run in production, it's an FDE role whatever it's called.
The skills that make a great FDE
I describe the effective FDE as π-shaped: broad working knowledge across the stack, with two deep legs.
- Broad engineering: Python or TypeScript, SQL, APIs, cloud, containers, security and identity basics, and enough Linux and networking to debug in someone else's environment.
- Deep leg one, data and AI: data platforms, retrieval-augmented generation, evaluation, and building and operating AI agents.
- Deep leg two, the customer: discovery, scoping, explaining trade-offs to non-engineers, and leading without formal authority.
Coding gets you the interview. The ability to turn a vague business problem into something buildable is what makes you valuable.
How to grow into the role
- Get strong at one engineering discipline first. Backend, data or platform engineering are the most common starting points.
- Seek out customer-facing work in your current job: requirements sessions, user interviews, production support, demos.
- Build AI applications end to end. Ship a retrieval or agent project that uses real documents and real tools, and evaluate how well it works.
- Practice writing and explaining. Short design docs and clear updates are a large part of the job.
- Learn to work in unfamiliar environments. Every customer's network, identity setup and data is different; comfort with ambiguity is the core trait.
Is it right for you?
You'll probably enjoy it if you like variety, travel or remote collaboration with customers, seeing your work used immediately, and solving the messy last mile of a problem. You may not if you prefer long, uninterrupted deep work on one codebase, or dislike context switching.
I'm writing a two-volume book on this role: The Forward Deployed Engineer. Volume 1 covers the foundations (software, cloud, data and AI); Volume 2 covers agents, the Model Context Protocol and the craft of working in the field.