Forward Deployed Engineers at Palantir, OpenAI and Anthropic: What They Actually Do

By Lokesh, Lead FDE Trainer at FDE Masters · Updated 7 October 2026 · 10 min read

A forward deployed engineer at Palantir, OpenAI or Anthropic writes production code inside a customer’s environment and owns the outcome. Palantir invented the role, OpenAI runs it as scoped three-phase engagements, and Anthropic’s version ships MCP servers, sub-agents and agent skills for enterprise accounts. Below: what each team does, what it pays, its India presence and how to prepare.

Where did the forward deployed engineer role come from?

A forward deployed engineer (FDE) is a software engineer embedded with a customer to integrate, customise and deploy a product until it delivers value. The term traces back to Palantir. If you want the full picture of the role before the company detail, start with what a forward deployed engineer is.

Palantir needed engineers who could make Gotham and Foundry work on messy government and enterprise data on site. By 2016, according to the Pragmatic Engineer’s history of the role, Palantir had more FDEs than product engineers. The model spread when ex-Palantir engineers joined AI labs in 2023 and 2024.

Andreessen Horowitz called FDE the hottest job in startups in June 2025. Postings grew 800% between January and September 2025 on Indeed, and 1,165% year on year in Bloomberry’s count (linked below). The job is no longer a Palantir oddity.

What do forward deployed engineers at Palantir do?

Palantir calls the role Forward Deployed Software Engineer (FDSE). An FDSE lives on a customer site, sometimes for months, and builds the data pipelines, ontologies and applications that make Foundry useful for that one customer.

The work is unglamorous by design. Day one might be reverse-engineering a 15-year-old SQL Server schema; week six might be training the customer’s analysts on the app you built from it. Product engineers back in Palo Alto take the patterns FDSEs find and turn them into platform features.

Public compensation data puts a US new-grad FDSE at $135K to 145K. Palantir hires new graduates into this track, which makes it the one FDE program at a top-tier company that is open to freshers, in the US at least. [PLACEHOLDER: verify whether Palantir currently hires FDSEs in India.]

What do forward deployed engineers at OpenAI do?

OpenAI’s FDE team runs customer engagements in three phases: scoping, where the team and the customer agree on a problem and a success metric; validation, where they build evals and prove the model can hit the metric; and delivery, where the FDE writes production code on the customer’s infrastructure. The Pragmatic Engineer piece linked above describes this structure in detail.

The team is small for the company’s revenue. An OpenAI executive told Business Insider the FDE group had 39 engineers and was planned to reach 52 by the end of 2025. Each engineer therefore carries large accounts, which is why the bar is senior.

What do forward deployed engineers at Anthropic do?

Anthropic’s FDE job description is the most concrete of the three. The role ships MCP (Model Context Protocol) servers that connect Claude to enterprise systems, sub-agents that split a workflow into supervised steps, and agent skills that package a repeatable task. It asks for 4+ years of experience and lists a base range of $280K to 320K.

The artefacts matter because they are the same ones the rest of the market is converging on. An Anthropic FDE might build an MCP server over a customer’s ServiceNow instance, wire a Claude Agent SDK workflow with a human approval step, and leave behind an eval suite that runs in the customer’s CI. Those three deliverables are now the core of most FDE curricula, including ours.

Anthropic opened a Bengaluru office in February 2026. Whether FDE roles are staffed from India is not public at the time of writing. [PLACEHOLDER: verify Anthropic India FDE openings.]

How do Databricks, Google Cloud and Scale run the role?

Databricks hires Senior FDEs in Bengaluru to build Lakehouse and agent solutions directly on customer data. Google Cloud has an Applied AI FDE function in Bengaluru focused on Gemini and Vertex deployments for large accounts. Scale AI uses FDEs to deploy its agent and evaluation platform into enterprise and government customers. [PLACEHOLDER: verify current Scale AI FDE openings and India presence.]

The common thread: the vendor sells a platform, the FDE makes it work for a named customer, and the FDE’s findings flow back to product. The main difference is seniority. Databricks and Google Cloud India postings skew to 5+ years, which is why most Hyderabad engineers reach them from a services FDE role rather than directly.

Company	What the FDE builds	Experience asked	Public pay signal	India presence
Palantir	Data pipelines, ontologies, Foundry and Gotham apps on site	New grad to senior (FDSE track)	US new-grad $135K to 145K	[PLACEHOLDER: verify]
OpenAI	Scoping, evals, production code on customer infra in 3 phases	Senior; team of 39 growing to 52	[PLACEHOLDER: not public]	Delhi office; Mumbai and Bengaluru planned
Anthropic	MCP servers, sub-agents, agent skills	4+ years	$280K to 320K	Bengaluru office since Feb 2026
Databricks	Lakehouse and agent solutions on customer data	Senior FDE	[PLACEHOLDER]	Bengaluru
Google Cloud	Applied AI deployments on Gemini and Vertex	Mid to senior	[PLACEHOLDER]	Bengaluru
Scale AI	Agent and eval platform deployments	[PLACEHOLDER]	[PLACEHOLDER]	[PLACEHOLDER: verify]

Do these companies hire forward deployed engineers in India?

Some do, mostly in Bengaluru. Google Cloud’s Applied AI FDE team and Databricks’ Senior FDE roles are both Bengaluru-based. Anthropic’s Bengaluru office opened in February 2026, and OpenAI has a Delhi office with Mumbai and Bengaluru expansion planned, but neither has publicly staffed FDE roles from India yet.

The wider Indian market is bigger than the labs. CIEL HR counted 52 organisations hiring FDEs in July 2026, up 130% year on year, with Bengaluru taking about half and Hyderabad in the top three. Sarvam AI, Cartesia, Turing, Sutherland, Pulsora, EPAM, TCS, Accenture and Coforge all use the title.

Pay follows the employer tier. CIEL HR’s clients pay ₹35 to 45 L at entry and ₹70 to 90 L at senior levels, Cartesia’s Bengaluru band is ₹70 to 90 L, while IT-services FDE practices pay ₹10 to 18 L. Our FDE salary in India guide has the full breakdown by source.

What do FDE job descriptions at these companies demand?

Read ten FDE postings and the same requirements repeat. Bloomberry’s analysis of 1,000 postings quantifies it: Python in 66%, TypeScript in 35%, AWS in 32%, AI agents in 35%, LLM in 31%. 55% list customer-facing work before any technology, and 68% mention travel.

• Shipped production systems, not notebooks. Every JD asks for things you deployed and operated.
• Customer communication. Running discovery, writing status updates for executives, demoing to non-engineers.
• LLM application skills. RAG, agents, tool calling, evals, prompt and context engineering, cost and latency control.
• Integration depth. SSO, RBAC, enterprise APIs (ServiceNow, Salesforce, SAP), message queues, data warehouses.
• Ambiguity tolerance. Phrases like “define the problem with the customer” appear in nearly every lab posting.

Notice what is missing: research credentials and model training. The role is applied engineering plus consulting. We list the full stack with learning order in our forward deployed engineer skills guide.

How should you prepare for each company?

Preparing for Palantir

Palantir’s loop tests decomposition under ambiguity more than LLM knowledge. Practise taking a vague business problem, splitting it into data, logic and interface, and explaining trade-offs out loud. Strong SQL, a systems-design round and a values conversation are standard; a deployed data application in your portfolio helps more than an AI demo.

Preparing for OpenAI

Expect to be asked how you would scope an engagement and what evals you would write before any code. Build one project where the eval suite exists before the feature, with a golden dataset and a pass threshold. Be ready to talk about a time you delivered code on infrastructure you did not control.

Preparing for Anthropic

Ship an MCP server against a real third-party system and publish it. Build a multi-step agent with the Claude Agent SDK that pauses for human approval before a write action. The job description names these artefacts, so interviewers will probe how you handled auth, failure modes and prompt injection in each.

Preparing for Databricks and Google Cloud

Both weight platform depth. For Databricks, know Spark, Delta tables, Unity Catalog permissions and how an agent reads from a governed lakehouse; for Google Cloud, know Vertex AI, IAM, VPC Service Controls and how to keep customer data inside a region. In both cases, one end-to-end deployment story beats a long list of certifications.

Whatever the company, the interview questions overlap heavily. Our FDE interview questions post collects the ones candidates report most often, with sample answers.

What is a realistic path to these teams from Hyderabad?

Very few engineers go straight from a Hyderabad college to an OpenAI or Anthropic FDE seat. The pattern that works is two steps: an FDE or solutions role at an Indian employer for 2 to 3 years, with real deployments and customer references, then a lab or platform company in Bengaluru or abroad.

That is how we structure training at FDE Masters. The FDE Advanced Program (10 weeks, ₹30,000 online or ₹35,000 classroom) is for engineers with 2 to 7 years who already code and want the MCP, agent, eval and deployment artefacts above on their resume, finished with a real 2-week forward deployment. Lokesh, our lead trainer, spent a year as an FDE at Brolly Software Solutions in Hyderabad deploying the RAG Chatbot and Brolly VoxFlow products into customer environments, which is the kind of story these interviews want to hear.

Frequently asked questions