What Is a Forward Deployed Engineer? The Complete Guide to AI's Hottest New Role
17.07.2026
CEO
Forward Deployed Engineer job postings are up 800% year-over-year. OpenAI, Anthropic, and Palantir, some of the biggest names in AI and data, are all hiring them. But most people still can't explain what they actually do.
Is it a consultant? A software engineer? A sales engineer?
It's all of them and none of them.
Here's what the role actually looks like, why it's growing so fast, and whether it's right for you.
Is it a consultant? A software engineer? A sales engineer?
It's all of them and none of them.
Here's what the role actually looks like, why it's growing so fast, and whether it's right for you.
What Is a Forward Deployed Engineer?
A Forward Deployed Engineer (FDE) is a software engineer who works directly with a customer to build, deploy, and operate custom software inside that customer's environment.
Think of it as half engineer, half consultant, full owner.
The role combines three things:
An FDE is not a sales engineer who demos products. They don't just write reports. They ship code. They stay accountable. And they work on the customer's turf with their messy data, their complex security, and their urgent real-world problems.
Think of it as half engineer, half consultant, full owner.
The role combines three things:
- Engineering is about writing production code. Real systems, real data, real customers.
- Consulting is about working directly with the customer, understanding their business, and translating vague problems into concrete solutions.
- Ownership means no handoffs. The engineer stays, operates, and owns the result.
An FDE is not a sales engineer who demos products. They don't just write reports. They ship code. They stay accountable. And they work on the customer's turf with their messy data, their complex security, and their urgent real-world problems.
Where Does the Term Come From?
The term was coined by Palantir, a data analytics pioneer, around 2010. They needed engineers on-site with customers because their software was too complex to deploy remotely. The traditional model – sell software, train integrators, exit – produced demos that wowed and deployments that stalled.
Palantir's answer was to embed engineers directly with customers. Those engineers wrote production code against the customer's actual data, sat in operational meetings, and stayed for the lifetime of the deployment.
The model worked. And as generative AI exploded in 2023-2024, the FDE role re-emerged as the answer to a familiar problem: AI products demo brilliantly and deploy unevenly.
Today, OpenAI, Anthropic, Ramp, and most modern AI companies maintain FDE teams.
Palantir's answer was to embed engineers directly with customers. Those engineers wrote production code against the customer's actual data, sat in operational meetings, and stayed for the lifetime of the deployment.
The model worked. And as generative AI exploded in 2023-2024, the FDE role re-emerged as the answer to a familiar problem: AI products demo brilliantly and deploy unevenly.
Today, OpenAI, Anthropic, Ramp, and most modern AI companies maintain FDE teams.
FDE vs Software Engineer vs Solutions Architect
The FDE gets confused with three adjacent roles. The distinctions matter.
The key difference: An FDE writes and owns production code in the customer's environment. No handoffs. No "let me check with the implementation team." The person who scopes the problem is the person who ships the solution and operates it.
What Does an FDE Actually Do?
Day-to-Day Responsibilities
An FDE's job spans the full arc of a customer deployment:
- Scoping. Meet with stakeholders from operators to VPs. Understand what problem actually matters. Often the customer can't yet articulate what's wrong. The FDE imposes structure on a vague situation without oversimplifying it.
- Building. Write production code. Integrations, data pipelines, backend services, RAG systems, agents, internal tools. Whatever blocks the deployment becomes the job.
- Deploying. The real work is navigating enterprise reality: SSO, legacy ETL pipelines, regulatory constraints, data residency, and the politics of getting production credentials from a customer's security team.
- Operating. When a deployment breaks, the FDE owns the fix. They diagnose and resolve it themselves, often while a customer team waits, instead of filing a ticket.
- Feeding back. The best FDEs spot patterns across deployments and push them back into the roadmap. If an FDE's work isn't influencing what gets built next, the role is being misused.
A Week in the Life of an FDE
A typical week for an FDE spans discovery, building, deploying, and incident response, often within the same five days. One day might be spent scoping data sources with the customer's team. The next could be a coding marathon building integrations. By the end of the week, they're responding to live issues and explaining technical decisions to non-technical stakeholders.
The work is messy, but it's close to revenue, adoption, and product truth. There's no guesswork about whether the work mattered – the customer environment tells you immediately.
The work is messy, but it's close to revenue, adoption, and product truth. There's no guesswork about whether the work mattered – the customer environment tells you immediately.
Why AI Companies Are Hiring FDEs
Agentic AI is structurally hostile to the traditional enterprise-software playbook. Three reasons FDEs are uniquely suited to it:
1. The agent's behavior is shaped by your data, not a settings menu.
A traditional SaaS product behaves the same for every customer. An AI system behaves differently because its outputs are shaped by your documents, your historical decisions, your exception logic, and your tone. Someone has to look at your actual data, write the prompts and tools and guardrails against it, and iterate. That's engineering work, not configuration work – and it doesn't end at deployment.
2. Edge cases are the work.
An agent that handles 95% of cases is not a deployable system – it's a science project. The last 5% – malformed PDFs, contradictory policy clauses, the customer who types in all caps is where the value sits, because that's where human operators spend their time. Capturing those edge cases requires sitting with operators, watching them work, and codifying their judgment into the agent. An FDE who is in your environment can do this in weeks. A remote vendor who hands off cannot do it at all.
3. The system has to be observed in production to actually work.
Agent quality is not a one-shot ship; it's an ongoing measurement. Did approval rates drift this week? Did a new email pattern start arriving that the agent doesn't handle? Did a downstream system change its API response format? These questions can only be answered by someone watching the live system with both engineering authority and operational context. That's the FDE's defining job after launch.
1. The agent's behavior is shaped by your data, not a settings menu.
A traditional SaaS product behaves the same for every customer. An AI system behaves differently because its outputs are shaped by your documents, your historical decisions, your exception logic, and your tone. Someone has to look at your actual data, write the prompts and tools and guardrails against it, and iterate. That's engineering work, not configuration work – and it doesn't end at deployment.
2. Edge cases are the work.
An agent that handles 95% of cases is not a deployable system – it's a science project. The last 5% – malformed PDFs, contradictory policy clauses, the customer who types in all caps is where the value sits, because that's where human operators spend their time. Capturing those edge cases requires sitting with operators, watching them work, and codifying their judgment into the agent. An FDE who is in your environment can do this in weeks. A remote vendor who hands off cannot do it at all.
3. The system has to be observed in production to actually work.
Agent quality is not a one-shot ship; it's an ongoing measurement. Did approval rates drift this week? Did a new email pattern start arriving that the agent doesn't handle? Did a downstream system change its API response format? These questions can only be answered by someone watching the live system with both engineering authority and operational context. That's the FDE's defining job after launch.
What Skills Does an FDE Need?
FDEs need a rare mix of skills: deep expertise in one core area and broad capability across several others.
Technical Skills
According to an analysis of 1,000 FDE job postings: Python (66%), AI agents (35%), TypeScript (35%), AWS (32%), and LLMs (31%).
Soft Skills
Who Should Hire an FDE?
When FDE Makes Sense
- Speed to market is critical
- The product is still evolving
- There's funding but no technical founder
- Agency bloat is a concern
- A partner is needed, not a vendor
When FDE Doesn't Make Sense
- Fixed scope
- Large team required
- Long timeline
- AI is your core business
How Much Does an FDE Cost?
FDE compensation is among the highest in tech, driven by the scarcity of engineers who are both strong builders and high-empathy communicators.
About 70% of FDE roles offer company shares in addition to salary. These shares can be worth more than the salary itself.
Frequently Asked Questions
A Forward Deployed Engineer embeds directly with a customer to scope, build, and operate custom software. They write production code, stay accountable, and own the result. Think half engineer, half consultant, full owner.
A consultant produces a deliverable and exits. An FDE ships production code and stays on it. The accountability sits with the same person who wrote it.
A solutions engineer supports the sales motion and rarely writes production code. An FDE joins post-sale and owns the implementation and operation.
AI products demo brilliantly but deploy unevenly. FDEs bridge the gap between generic AI capability and specific customer problems – building the actual integrations, guardrails, and workflows that make AI work in production.
Entry-level: $140K–$250K. Mid-level: $200K–$300K. Senior at AI labs (OpenAI/Anthropic): $350K–$550K. Staff-level: $630K+.
Strong Python, SQL, cloud experience, and AI fluency (RAG, agents, evals). Plus customer-facing skills: empathy, radical ownership, problem decomposition, and communication under pressure.
Projects that require speed, flexibility, and deep customer integration. If you need to launch in weeks, adapt quickly to feedback, and don't have a technical founder — an FDE is often the right choice.
Ready to Explore If an FDE Is Right for You?
Forward Deployed Engineers are becoming the backbone of how AI gets deployed in the real world. They bridge the gap between what AI can do and what actually works for customers.
If you're a founder or CTO considering this model, or if you're exploring the role for yourself, the first step is understanding whether FDE is the right fit for your situation.
Book a free 20-min call with an MXA AI architect
If you're a founder or CTO considering this model, or if you're exploring the role for yourself, the first step is understanding whether FDE is the right fit for your situation.
Book a free 20-min call with an MXA AI architect