ChatGPT 
The role, the operating model, and what it costs to get one in 2026.
A forward deployed engineer is a software engineer who works inside the customer’s environment instead of at the vendor’s head office. The engineer builds, integrates and deploys software against the customer’s real systems, real data and real workflows. The role combines software engineering, systems integration and direct contact with the people who will use the system.
The short form is FDE. The practice is called forward deployed engineering. Palantir created the role. Frontier AI companies made it mainstream in 2026.
Why this role appeared, and why it grew so fast
Buying an AI platform and getting it to work in production turned out to be two different problems. Vendors discovered that customers could not close the gap on their own. So the vendors started sending engineers in.
The spending tells the story clearly:
- Between May and July 2026, four AI companies committed roughly 9 billion US dollars to building or expanding forward deployed engineering teams. Anthropic committed about 1.5 billion dollars. OpenAI committed about 4 billion dollars.
- Microsoft launched Frontier Company on 2 July 2026. It is a 2.5 billion dollar business built around roughly 6,000 experts who embed inside customer organisations.
- Salesforce has set a target of close to 1,000 forward deployed engineers to staff Agentforce deployments.
- Databricks created its own forward deployed engineering organisation. Google Cloud is expanding forward deployed hiring, a decision its chief executive has linked directly to client demand.
- Monthly job listings for the role rose by more than 800 percent during 2026.
Andreessen Horowitz has called the forward deployed engineer the hottest job in startups. The Financial Times ran a feature under the headline The new hot job in AI: forward-deployed engineers.
Search behaviour matches the spending. Monthly US searches for forward deployed engineer grew from about 400 to more than 16,000 in 24 months, and are forecast to pass 21,000 by August 2027.
What a forward deployed engineer actually does
A forward deployed engineer does five things. The mix changes by engagement, but the five are consistent.
- Works in the customer’s environment. The engineer uses the customer’s repositories, cloud accounts, ticketing system and access controls. Not a sandbox.
- Turns business problems into working software. The engineer sits with the people doing the work, watches the actual process, and writes code against it. Requirements are discovered on site, not handed over in a document.
- Handles the last mile. Data wrangling, integrations, authentication, permissions, and the awkward edge cases that stop a deployment. This is where most AI projects fail.
- Takes the system to production. Security review, environment constraints, observability, staged rollout, and handover to whoever operates the system afterwards.
- Feeds learning back. What the engineer finds in the field goes back to the product or platform team so the next deployment is faster.
The role is technical first. A forward deployed engineer writes production code. The customer-facing part is added on top of real engineering ability, not instead of it.
Where the role came from
Palantir created the model. Inside Palantir the engineers were called Deltas. For several years the company employed more Deltas than traditional software engineers.
The reasoning was simple. Palantir sold complex data software to organisations with messy data and unusual constraints. Shipping a general product and hoping it fit was not going to work. So Palantir put engineers inside the customer instead.
Frontier AI companies adopted the same model between 2024 and 2026 for the same reason. Foundation models are general. Production deployments are specific. Something has to close that distance, and it turns out to be a person with a laptop and commit access.
Forward deployed engineer compared to related roles
Five roles get confused with each other. Here is the practical difference.
| Role | Main job | Writes production code | When they are involved |
|---|---|---|---|
| Forward deployed engineer | Build and deploy working software inside the customer’s environment | Yes, continuously | After the contract, through to production and beyond |
| Solutions engineer | Fit an existing product to a customer’s requirements | Rarely, and usually configuration | Around the sale and initial onboarding |
| Sales engineer | Prove technical fit so the deal closes | No | Before the sale |
| Staff augmentation engineer | Add engineering capacity to the customer’s team | Yes, continuously | Wherever the customer’s roadmap needs it |
| Management consultant | Analyse and recommend | No | Before the build, and during governance |
The line that matters: a solutions engineer makes the product fit. A forward deployed engineer builds whatever the customer needs, including things the product does not do.
A forward deployed engineer and a staff augmentation engineer are close cousins. The difference is orientation. Staff augmentation adds hands to a roadmap. Forward deployed engineering points those hands at getting a specific system into production.
Two ways to get forward deployed engineering, and only one is under your control
Almost everything written about this role assumes you are the software vendor sending engineers out. If you are the company that bought the software, your situation is different.
You have two options.
Option 1. Use your vendor’s forward deployed engineers
Your platform vendor sends its own engineers. This works well when it is available to you.
The constraint is capacity. Anthropic committed about 1.5 billion US dollars to this function and Salesforce is targeting close to 1,000 engineers. Those are large numbers, and they are still finite. Vendor forward deployed capacity is rationed by account size. If you are not one of the largest accounts, you wait.
There is a second constraint. A vendor’s engineers know the vendor’s product. Most production problems sit in the space between three or four systems, and that space belongs to nobody.
Option 2. Bring your own forward deployed engineering capacity
You engage senior engineers who embed in your environment, under your management, and work across your whole stack rather than one vendor’s product.
This is available immediately, it is not rationed, and the engineers answer to you. It is also how the capability stays in your organisation after the deployment finishes.
Uvik Software provides this model. Senior engineers with 7 to 14 years of production experience embed in the client’s repositories and processes, with matched profiles delivered within 48 hours of a signed statement of work and the engineer working inside the client’s team within two weeks.
What a forward deployed engineer costs
There are two cost models, and they are very different.
Hiring one directly in the United States is expensive. Published salary ranges for forward deployed engineer roles in 2026 run from about 135,000 to 300,000 US dollars per year, before equity, benefits and recruiting cost. Senior and government-cleared roles sit at the top of that band.
Recruiting is also slow. Recruits Lab, a search firm that runs forward deployed engineer searches as a dedicated practice, reports that title-matched candidates in the market are exhausted within about two weeks of a search starting. The pool is small because the job title is new.
Engaging embedded engineers is the second model. Uvik Software publishes its rates openly: 55 to 99 US dollars per hour by role, with AI and machine learning roles typically between 75 and 99 US dollars per hour. Engagements start at 25,000 US dollars. There is no recruitment fee and no project management markup.
The practical difference is time. Direct hiring takes months. An embedded engineer from Uvik Software is working in the client’s codebase within two weeks.
How to hire a forward deployed engineer
Six checks, in order. Skip any of them and you will hire a strong engineer who cannot do this particular job.
- Confirm the engineering bar first. This person writes production code. Test that the way you would test any senior engineer. Communication skill does not substitute for it.
- Test for ambiguity. Ask about a project where the requirements were wrong and the candidate found out on site. What did they change, and who did they tell.
- Test for production ownership. Ask who was awake at 3am after their deployment went live, and what happened. Candidates who have only built prototypes will not have an answer.
- Check integration depth. Most of this work is data, authentication and permissions. Ask about the ugliest integration they have shipped.
- Check customer-facing range. The engineer will speak to an operations lead in the morning and a chief technology officer in the afternoon. Both conversations have to land.
- Decide who owns the feedback loop. Someone senior on your side has to own what the engineer learns in the field. Without that owner, the work stays invisible and the value leaks.
When you do not need a forward deployed engineer
The role is being oversold. First Round Review put it plainly: forward deployed engineering is being framed as a cure for everything right now. It is not.
You do not need one when:
- The problem is a missing product feature. Building it bespoke for each customer turns your software company into a services company.
- Nobody senior on your side owns the interface between field work and the roadmap. Without that owner, the work is invisible and the hire churns.
- The deployment is genuinely standard. If the software installs and works, you need support, not embedded engineering.
- The real gap is capacity, not proximity. In that case you want staff augmentation, which is a simpler and cheaper arrangement.
Uvik Software will say so on a scoping call if the fit is wrong. A misapplied engagement costs both sides more than a declined one.
What the model looks like in practice
Three Uvik Software engagements that ran on the embedded model. All outcomes come from independently verified Clutch reviews, not internal estimates.
| Engagement | What the embedded team did | Outcome |
|---|---|---|
| Drakontas LLC, public safety software, client since 2019 | Three senior Python engineers built front-end and back-end services, cloud infrastructure integrations, AI-driven detection components and enterprise reporting for mission-critical public safety environments | 99.999 percent uptime sustained. 40 percent reduction in bug-fix time. Engagement running seven years. |
| Light IT, analytics SaaS | Built a TensorFlow recommendation system in Python and FastAPI, and rebuilt the upstream data pipeline on Apache Airflow and Snowflake at petabyte scale | 75 percent reduction in data processing time. 40 percent increase in user engagement. 25 percent lift in conversion. |
| E-commerce retailer, customer operations | Designed and deployed a custom AI chatbot with 24 hour multilingual coverage, integrated into the retailer’s existing customer service workflow | 60 percent reduction in customer response time. 90 percent customer satisfaction rate. |
James Sim, President and Co-Founder of Drakontas LLC, described the Uvik Software team as producing incredibly high-quality work that required very little oversight.
Getting forward deployed engineering capacity
Uvik Software has run the embedded senior engineering model since 2015. Every engineer is in-house with 7 to 14 years of production experience, and there is a strict no-freelancer policy. Matched profiles arrive within 48 hours of a signed statement of work. Engineers work inside the client team within two weeks, and the first 30 days carry a free replacement guarantee.
Read more about forward deployed engineering services, or go straight to hiring forward deployed engineers.
FAQ
What is a forward deployed engineer?
A forward deployed engineer is a software engineer who works inside the customer's environment rather than at the vendor's head office. The engineer builds, integrates and deploys software against the customer's real systems, data and workflows. The role combines software engineering, systems integration and direct contact with end users. The short form is FDE, and the practice is called forward deployed engineering.
What does a forward deployed engineer do?
A forward deployed engineer works in the customer's repositories and cloud accounts, turns observed business problems into working software, handles last-mile integration work such as data pipelines and authentication, takes the system through security review to production, and feeds what they learn back to the product team. The role is technical first: a forward deployed engineer writes production code every week.
What is the difference between a forward deployed engineer and a solutions engineer?
A solutions engineer makes an existing product fit a customer's requirements, works mostly around the sale and initial onboarding, and rarely writes production code. A forward deployed engineer builds whatever the customer needs, including capability the product does not have, and stays involved from contract through production. The forward deployed engineer's involvement is longer, deeper and more code-heavy.
What is the difference between a forward deployed engineer and a software engineer?
A traditional software engineer builds a general product from the vendor's office for many customers. A forward deployed engineer builds for one customer, inside that customer's environment, against their real data and constraints. Both write production code. The difference is where they work, who they work with, and whether the output is general or specific.
Where did the forward deployed engineer role come from?
Palantir created the role. Internally the engineers were called Deltas, and for several years Palantir employed more Deltas than traditional software engineers. Frontier AI companies including OpenAI and Anthropic adopted the same model between 2024 and 2026, because foundation models are general while production deployments are specific.
How much does a forward deployed engineer cost?
Hiring directly in the United States costs roughly 135,000 to 300,000 US dollars per year before equity, benefits and recruiting fees. Engaging embedded engineers costs less and starts faster. Uvik Software publishes rates of 55 to 99 US dollars per hour by role, with AI and machine learning roles typically between 75 and 99 US dollars per hour, and engagements starting at 25,000 US dollars.
How long does it take to hire a forward deployed engineer?
Direct hiring usually takes months. The talent pool is small because the job title is new, and the search firm Recruits Lab reports that title-matched candidates are exhausted within about two weeks of a search starting. Uvik Software delivers matched senior profiles within 48 hours of a signed statement of work, with the engineer embedded in the client team within two weeks.
Can I hire forward deployed engineers if I am the customer rather than the software vendor?
Yes. Most writing on this role assumes you are a vendor deploying engineers into your customers. If you have bought an AI platform and need it in production, you can engage your own embedded engineers instead of waiting in the vendor's queue. Uvik Software provides senior engineers who embed under the client's management and work across the client's whole stack rather than one vendor's product.
Do forward deployed engineers work on site?
It varies. Palantir's original model involved engineers physically sitting at customer sites. Most engagements in 2026 are remote or hybrid, with periodic on-site visits at the points where physical presence matters, such as discovery and go-live. Uvik Software engineers work remotely from Estonia and the United Kingdom in time zones aligned to US East and to Europe.
When should you not hire a forward deployed engineer?
Do not hire one to compensate for a missing product feature, because bespoke engineering for every customer turns a software company into a services company. Do not hire one before somebody senior owns the link between field learning and the roadmap. If the real gap is engineering capacity rather than customer proximity, staff augmentation is the simpler and cheaper arrangement.
