Forward deployed AI engineer: the job, the work and who hires

We read these employers' job postings and announcements on 2 October 2026. Every quote below is their own wording, and every number names its source.

Why AI companies are hiring forward deployed engineers now

OpenAI and Anthropic now build whole companies around this work. On 11 May 2026 OpenAI launched the OpenAI Deployment Company, which "will extend OpenAI's ability to embed engineers specialized in frontier AI deployment, known as Forward Deployed Engineers, or FDEs, into organizations". It launches with "more than $4 billion of initial investment" and is "majority-owned and controlled by OpenAI". Axios reported its valuation at $14 billion; OpenAI's announcement gives none. OpenAI also "has agreed to acquire Tomoro, an applied AI consulting and engineering firm", with "approximately 150 experienced Forward Deployed Engineers and Deployment Specialists".

A week earlier, on 4 May, Anthropic announced an enterprise AI services firm for mid-sized companies with Blackstone, Hellman & Friedman and Goldman Sachs. "Applied AI engineers from Anthropic will work alongside the firm's engineering team", the announcement says. Goldman Sachs said the new firm would help by "democratizing access to forward-deployed engineers". The firm announced the acquisition of Fractional AI as its operational core in May and launched as Ode with Anthropic in July.

The labs hire FDEs directly too. On 2 October 2026 OpenAI's careers board listed 24 postings with "Forward Deployed" in the title, from San Francisco to Seoul and Madrid. Anthropic's London posting calls the hire "one of our founding FDEs". Salesforce wrote in October 2025 that it is "laser-focused on scaling one of our most exciting roles: the Forward Deployed Engineer (FDE)". Google Cloud hires FDEs for GenAI and Applied AI, and Cohere had 15 FDE-titled postings open on the day we read its board.

Search demand grew too. Google's Keyword Planner, our read on 2 October 2026, estimates searches for "forward deployed engineer" at 18,100 worldwide in September 2025 and 165,000 in August 2026, with a 12-month average of 74,000. Searches for "forward deployed engineer course" went from 0 to 5,400 between the same two months.

Better vendor tools add to this work. The launches at OpenAI's DevDay 2026 still left companies with agent rules, a data map, model costs and integrations to set up on their own. That setup happens at the customer, and it is forward deployed work.

The work on an LLM project, in employers' own words

Put the AI postings side by side and the same loop appears. OpenAI's San Francisco FDE posting names it in one sentence: "You will own discovery, technical scoping, system design, build, and production rollout".

Discovery sets the target. Cohere asks its FDEs to turn "ambiguous business problems into well-framed agentic workflows with clear success criteria and evaluation methodologies". You agree on what success means before you write the first prompt.

The build happens inside the customer's systems. Anthropic's FDEs "work within customer systems to build production applications with Claude models". Much of the engineering is the connection: the agent reads the customer's data, calls the customer's tools and works within the customer's permissions.

Evals decide what ships. Cohere wants "robust evaluation frameworks, moving well beyond trial and error, to measure agent accuracy, safety, and latency". Without an eval set, nobody can measure what the last prompt change did to the agent.

Adoption is the scorecard. OpenAI's Applied AI Engineer posting for India says success "is measured by production systems, sustained adoption, and meaningful customer impact". A pilot that impresses in a demo and then sits unused fails that test. I mapped the stages where AI projects stall between a working pilot and production, and getting through them is what FDEs are paid for.

Security comes with the build. OpenAI hires a separate Forward Deployed Security Engineer for its deployments. Cohere wants every agent "reliable, observable, safe, and auditable from day one".

Field patterns go back into the product. Anthropic asks its FDEs to "Identify and codify repeatable deployment patterns". Google turns field work into "reusable modules or product feature requests". The tenth customer gets the building blocks of the first nine.

Applied AI engineer, forward deployed engineer, Frontier Deployed Engineer: reading the titles

The work is consistent. The titles vary, and they decide which jobs your search finds.

Forward deployed engineer comes from Palantir, where these engineers are called "Deltas" and focus on "one customer, many capabilities". Palantir also hires under the title Forward Deployed AI Engineer today. The role reaches beyond AI, and I wrote up the full definition, from Palantir's Deltas to today's postings separately.

Applied AI engineer is the second title to search for. Anthropic lists far more Applied AI roles than FDE roles, and its FDE posting opens with "As a member of the Applied AI team at Anthropic, you will be a Forward Deployed Engineer (FDE)". OpenAI's India role is an "Applied AI Engineer", and its posting lists forward-deployed engineering among the relevant backgrounds. Mistral puts both terms in one title: "Applied AI, Technical Lead, Forward Deployed AI Engineer". Search for both, and read the duties: Anthropic's Applied AI Architect is a pre-sales role.

Frontier Deployed Engineer is a training track. On 2 October 2026 Anthropic announced the Claude Frontier Academy, which "aims to train 10,000 Frontier Deployed Engineers (FDEs) by the end of 2027", backed by $100 million. Same acronym, different first word. Participation is by nomination: organisations ask their Anthropic account team or Partner Account Manager whether they are eligible. Passing a multi-day in-person program earns the Claude Resident Engineer badge. Then comes a 12-week residency on a real Claude use case at the engineer's own organisation, and a final assessment for the Claude Frontier Deployed Engineer badge, the first expected in early 2027. Anthropic issues both badges, and nomination is the only way in.

Infosys adds one more term, Frontier Engineers. It appears in the Indian numbers below.

India: what the FDE numbers count

Indian headlines about FDE hiring run ahead of their sources. Here is what each one says.

CIEL HR: 130%. CIEL HR's August 2026 study says FDE hiring in India "has grown 130% over the past year". The method: AI deployment job openings at 52 companies, counting FDE roles together with "closely related AI deployment positions", with no absolute counts and no baseline published.

TCS: at least 1% of the workforce. On the earnings call of 9 July 2026, COO Aarthi Subramanian set a "target to have at least 1% of our employee base, to work in the new operating model", adding that "it is a transition". TCS ended that quarter with 593,798 employees, so 1% is about 5,900 people (our arithmetic), moving from the existing workforce into the new model.

Infosys: 6,000 Frontier Engineers. CEO Salil Parekh on 23 July 2026: "Our plan is to have 6,000 Frontier Engineers over the next few years." Asked about Forward Deployed Engineers, he gave the same figure. It is a target for the team's size, and the transcript attaches no hiring number to it.

OpenAI and Anthropic: Applied AI titles. On 2 October 2026 OpenAI's India openings included an Applied AI Engineer in Delhi, Mumbai and Bangalore, and none of its 24 FDE-titled postings was in India. Anthropic opened its Bengaluru office in February 2026 to "offer applied AI expertise to enterprise customers, digital natives, and startups". Its India postings include Applied AI Architect roles in Bangalore and Mumbai, and none is titled FDE. In India, the lab jobs closest to forward deployed work carry Applied AI titles.

The skills AI employers ask for

We read eight current FDE postings in full on 2 October 2026: Palantir, OpenAI, Anthropic, Scale AI (its public sector role), Ramp, Salesforce, Cohere and Google. All eight describe customer-facing work. Seven ask for LLM or agent experience, seven name Python and five name evals.

• Production LLM experience. "Production experience with LLMs including advanced prompt engineering, agent development, evaluation frameworks, and deployment at scale" (Anthropic).
• Agent development. MCP servers, sub-agents and agent skills at Anthropic; "Extensive experience building with and for agents" at Ramp.
• Evaluation frameworks. Golden datasets, regression suites and LLM-as-a-Judge, as in the table above.
• Python. "5 years of experience with software development using Python or similar coding languages" (Google).
• Customer-facing delivery. A hard requirement at Salesforce, with travel from 20% to 50% of the time across the postings.

OpenAI and Google ask for five years of experience, Anthropic for four, Ramp prefers three, and Palantir's general FDSE posting asks for one year after college.

If you write software today, the gap to close is production LLM work and time in front of customers. The routes employers describe, and the order in which to build these skills, are in the step-by-step path from software engineering into forward deployed work.

Start this week with one process at your current job. Build an agent for it in the tools your team already uses, write the eval set before you tune the prompt, and check a month later whether people still use it.

Frequently asked questions