How Palantir Invented the Forward Deployed Engineer Model and Why AI Startups Are Adopting It

In the early 2000s, Palantir had a problem that most software companies never had to solve. Its customers were intelligence agencies. These organisations were unable to clearly explain what they needed. They were not able to openly share data. Even their own workflows were in a constant state of change.

This made traditional product development almost useless. There were no clear requirements, no proper feedback loops and no way to run normal user interviews.

So Palantir did something unusual. Instead of asking customers what they wanted, it put engineers directly inside customer environments. These engineers learned by observing, experimenting and building in real time.

That decision resulted in the forward-deployed engineer model. What began as a necessity later became a great strategy. Today, it is being adopted by many AI companies for similar reasons.

Why Palantir Needed to Invent a New Engineering Model

The forward-deployed engineer origin is inextricably linked to the nature of Palantir's early work. Intelligence agencies are involved with sensitive and complex issues. These problems are not completely specified in advance.

In most software companies teams are based on clear problem statements. They collect requirements, design solutions and then construct products. Palantir was not able to follow this process.

The company realised something important. For complex and new problems, customers often do not know what they need until they see it working. This meant product discovery had to take place within the real environment, not outside.

To solve this, Palantir embedded engineers with customers. These engineers worked closely with analysts and operators. They observed the way work was done and constructed tools to make it better.

This approach was the basis of the forward-deployed engineer model.

Over the years, it scaled throughout the company. Until around 2016 Palantir had more forward-deployed engineers than traditional software engineers. This demonstrates the centrality of the model.

Later, with the launch of Foundry in 2016, Palantir began standardising its products. Many FDEs migrated to core engineering positions. But the model itself was still critical for the solution of complex problems.

The Echo and Delta Team Structure

A key part of forward-deployed engineering at Palantir is its team structure. Palantir did not depend on individual engineers working alone. It developed a system that was based on Echo and Delta teams.

Echo teams are experts in their domain. They are often from the same industries as the customers. For example, military, healthcare or finance.

Their role is to know real problems. They see where technology can create value. They also serve as a bridge between the customer and the engineering teams.

Echo team members are often considered people who challenge existing systems. They know what is broken and they want to fix it.

Delta teams are engineers who are execution-focused. They build solutions fast. They prioritise speed and impact over perfect design.

These engineers are comfortable working in uncertain conditions. They are dealing with incomplete data and changing requirements.

The strength of this model lies in the collaboration between the two teams. Echo teams identify the right problems and Delta teams build the solutions.

Together, they serve as a mini startup within each client environment.

Palantir explains this well:

You can think of a Dev's focus as one capability for many customers, while a Delta's focus is many capabilities for one customer.

This defines the Palantir forward-deployed engineer approach.

The Gravel Road to Paved Highway Feedback Loop

One of the most important ideas in the forward-deployed engineer model is the feedback loop between field work and product development.

This is referred to as the gravel road to paved highway FDE concept.

Forward-deployed engineers first develop rough solutions for specific customers. These are the "gravel roads." They are fast, pragmatic and focused on solving immediate problems.

Then, the core engineering team is responsible for studying these solutions. They find patterns among various customers. On the basis of this, they build standard features. These become the "paved highways."

This loop enables Palantir to scale custom work into product capabilities.

This is extremely different from consulting. Consultants are often one-time solutions and they move on. Their work does not typically become part of a larger product.

FDEs work differently. Every solution they build is a contribution to future product development.

However, this system requires discipline. If not managed well, it can become pure custom work which is hard to scale. Palantir was successful because it struck a balance.

FDE Model vs Traditional SaaS Delivery

The FDE model and traditional SaaS are two fundamentally different ways of delivering software to enterprise customers.

SaaS is built for scale. One product serves many customers with minimal customisation. Growth comes from self-serve adoption, and value builds gradually as users learn the product. It works best when the problem is well understood and the market is consistent.

The FDE model works in the opposite direction. Engineers go to the customer, not the other way around. Solutions are built specifically for each client environment. Sales happens top-down to executives, and the pitch is about outcomes rather than features. Value is delivered fast through working prototypes, not through gradual onboarding.

Neither model is better than the other. They solve different problems at different stages. SaaS wins when the market is ready for a standardised product. FDE wins when the problem is still being defined.

Why the AI Era Has Resurrected and Extended This Model

The emergence of AI has set the ideal conditions for the FDE model AI startups trend.

AI systems are not simple products. They rely on data, infrastructure and real-world conditions. These factors differ from organisation to organisation.

This means that no one solution works everywhere.

Many companies also have difficulty in defining what they want from AI. They need help with understanding use cases for building solutions.

This is exactly the type of problem the forward deployed engineer model is meant to solve.

OpenAI is a strong example. In early 2025, it established its FDE team headed by Colin Jarvis. The team began small but soon grew in size and spread to multiple cities and continents.

These engineers work directly with customers. They write code on customer systems and contribute to the product development.

They also worked on real-world deployments. One example is the John Deere project in Iowa. Engineers worked with farmers to create artificial intelligence-based crop solutions.

This demonstrates in practice how forward deployed engineer OpenAI teams work.

Ramp is another example. It implemented its FDE function with about 15 engineers organised in pods. This helped in improving the enterprise customer outcomes.

The trend is growing fast. Andreessen Horowitz has called FDE one of the hottest jobs in tech.

The reason is simple. AI problems are complex, undefined and highly customised. This makes the FDE model very effective.

What This Means to Engineers Considering the FDE Path

The development of the forward deployed engineer model reflects a long-term change in the industry. This is not a short trend. It is a reaction to the complexity that modern technology has reached.

Ambiguity is something that engineers in this role must deal with. They need to be able to work across systems and understand business problems and build practical solutions quickly.

As more and more companies implement this model, the demand for such engineers will only increase.

Understanding the origin of the forward deployed engineer is helpful in understanding why this role exists and is important today.

For people who are looking in this direction, structured learning platforms such as FDE Academy offer a starting point to develop the necessary mindset and skills.

TL;DR

The Forward Deployed Engineer model was invented by Palantir in the early 2010s as a response to working with intelligence agency customers who could not disclose their needs through normal product discovery.

• Palantir embedded engineers directly at customer sites, a model that grew until FDEs outnumbered traditional software engineers by 2016
• Echo teams bring domain expertise from the client's own field and identify the highest-value problems
• Delta teams are rapid prototyping engineers who build solutions quickly under imperfect conditions
• The gravel road to paved highway loop converts custom FDE field work into standardised product features
• The FDE model differs from SaaS in customer interaction, sales approach, customisation depth, and time to value
• AI companies including OpenAI and Ramp have adopted this model because AI deployment faces the same undefined, heterogeneous challenges Palantir faced in 2003
• a16z has called FDE the hottest job in tech, reflecting how central this model has become to AI product delivery