Forward Deployed Engineer in Finance: Building AI Inside Banks and Funds

A forward deployed engineer in finance is a software engineer who works inside a bank, a fund or a company a fund owns, building AI and data systems that run on that institution's own data. Most of these engineers are still sent in by the firms selling the technology: the AI labs, the data platforms and the large consultancies.

By September 2026, some of the biggest banks were advertising the title for their own staff, and a buyout firm had advertised for its own team to work inside the companies it owns.

For a candidate, that split matters more than the title. Whose badge you wear decides who your customer is, how long you stay, and what you are expected to leave behind. The work itself rewards two things: production code, and the finance habit of proving that every number is right.

25 to 50%

Time on the road most vendor postings ask for

2026 postings from labs, platforms and consultancies

8 to 24 weeks

One embedded engagement

as a buyout firm’s 2026 posting set it

$4bn+

Initial investment in one lab’s new deployment company

announced May 2026

What is a forward deployed engineer?

A forward deployed engineer takes a product that already exists and makes it work inside one customer's operation: that customer's data, systems, rules and people. The code is production code, but the engineer is measured by the customer's result rather than by what the product can do in general.

Where the title comes from

Palantir, the data-software company, still calls the role "the blueprint" in its job postings: "We pioneered this unique position, embedding talented engineers directly with our customers to tackle their most pressing challenges head-on."

In a 2019 post on Palantir's blog, a hiring manager explained that the company's two largest engineering roles by headcount were known internally as "Devs" and "Deltas." Delta was "a throwback to our early days, when each team in Business Development was named after a letter in the NATO alphabet." The post put the difference in one line: "You can think of a Dev's focus as 'one capability, many customers,' while a Delta's focus is 'one customer, many capabilities.'"

As the title spread beyond Palantir, its meaning stretched. Vinoo Ganesh, who led Palantir's program for turning its software engineers into forward deployed engineers and later ran business engineering at Citadel, described a dinner of people who hold the job where "we were all using the same two words (forward deployed) to describe jobs that had almost nothing in common."

His own definition keeps the product at the center: "An FDE solves customer problems in order to earn the insight that informs what gets built next." A team that fixes a customer's last mile but never sends the lesson home is, in his words, "a services/consulting team with a better title."

What does a forward deployed engineer do day to day?

In finance, earning that insight starts with learning how the institution actually runs, which is rarely how its documents say it runs.

Learning the nouns and verbs

Ganesh's shorthand for that first stretch is that the engineer's job is to "collect nouns and verbs." The nouns are the things a business treats as real: "A position, or a trade, or a counterparty." The verbs are how those things move: "how a trade gets booked, or what has to be true before the books can close, or who signs off on an exception at eleven at night and what happens when that person is on vacation."

Very little of that is written down. It lives in the heads of long-serving staff and in spreadsheets that whole teams quietly depend on, so the engineer sits with the people doing the work instead of reading about it.

Finance adds one constraint on top. The company Ganesh co-founded sells to "hedge funds, investment banks, PE firms, and other financial institutions," and he wrote that all of them share "a single non-negotiable: numbers have to be right, and someone has to be able to show why they are right." Every figure a finance deployment produces needs a trail back to its source, and building that trail is part of the job.

Building inside the bank

JPMorgan's International Private Bank described the building half in a July 2026 posting for an AI solutions engineer, forward deployed. The engineer would be "deployed directly into business contexts to identify where AI can create value," embedded "with IPB advisors, business teams, and product partners."

The job was to prototype fast, harden the result with the bank's core AI team, and own it through "hand-off to production," inside the team's "guardrails, evaluation, observability." That loop, repeated, is the week: find a problem worth solving, build something that works on the bank's own data, show that it is right, and hand it over.

What the first year looks like

String those loops together and a first year is counted in engagements rather than tickets. The clearest public description of a good one came from a buyout firm. Vista Equity Partners' 2026 posting for a senior forward deployed engineer described embedding "inside the engineering organizations" of its more than 90 enterprise software companies for "focused 8 to 24 week engagements," owning delivery "from first discovery call through operationalization and handoff."

It then set out what a strong engineer should have done "By month six":

1. Shipped at least one engagement that measurably cut the effort of keeping existing systems running or sped up feature delivery, "with metrics the company's engineering leadership signs off on."
2. "Left behind a working set of agentic workflows, evals, and tooling the company's engineers operate without external support."
3. Contributed at least one reusable accelerator or playbook "the next FDE engagement uses on day one."
4. "Built credibility with senior engineering leadership at the company, to the point of being requested back."

The second item is the one that defines the job: an engagement ends when the customer can run what was built without the engineer in the room.

An engagement, week by week

A twelve-week engagement with a bank's financial-crimes investigators could run like this, from the first interviews to a system the bank runs on its own.

Who hires forward deployed engineers in finance?

Those engagements are staffed from three places: the vendors that sell AI and data software to finance, the banks themselves, and on the buy side, a buyout firm working on its own portfolio. These are the clearest recent postings for each seat, with the base pay each employer published.

The vendor seat

Most forward deployed engineers working in finance sit here. OpenAI's financial-services posting, since closed, described "partnering with banks, asset managers, and private capital investors to deploy next-generation AI capabilities across their operations, investment processes, and portfolio companies," and leading the early build of the lab's industry offering for the sector.

Anthropic's engineers have done that work inside named institutions. Goldman Sachs's chief information officer, Marco Argenti, told CNBC in February 2026 that the bank had spent six months working with embedded Anthropic engineers to build autonomous agents for trade accounting and for client vetting and onboarding. The rest of that program is on the Goldman Sachs AI page.

In May 2026, FIS, the banking-technology company, said Anthropic's forward deployed engineers were embedded with it to co-design its financial-crimes agent and to establish the evaluation frameworks and knowledge transfer behind the rest of its agent roadmap. The engineers sat with FIS rather than inside each bank; BMO, FIS said, would be among the first banks to deploy the agent.

A lab also splits an engagement between roles. At Anthropic:

• the forward deployed engineer owns the build: "FDEs own the technical artifact"
• a technical deployment lead owns "the engagement end-to-end," from the statement of work to production
• pre-sales solutions architects, working with account executives, help sell the product in the first place, to financial-services customers across "banking, insurance, asset management, and payment providers"

The data platforms and consultancies fill out the vendor seat. Databricks runs its own AI forward deployed engineering team, "a highly specialized customer-facing AI team," and Deloitte hires forward deployed engineers to "embed with clients" on the Databricks platform. Vendor engineers spend much of their time at customer sites: most of these postings ask for a quarter to a half of working time on the road.

The bank's own seat

Banks are the newer employers of the title. Morgan Stanley advertised in September 2026 for a forward deployed engineer in Research Technology, "to expand and accelerate AI initiatives in the Research department," building "Search and Distribution Systems" for the firm's research and taking "Proof of Concepts" through to deployment. The bank's wider AI work is on the Morgan Stanley AI page.

Citi advertised in August 2026 for forward deployed engineers in AI transformation, working with business process owners "to fundamentally reimagine and re-engineer core business processes," and asked them to "apprentice future experts, ensuring the onward sustainability and scalability of our AI capabilities." A second Citi version, an AI business analyst working "as a Forward Deployed Engineer," sits between business users and the bank's data scientists. Citi's wider AI program is on the Citi and Bank of America page.

JPMorgan has used the title in its private bank, for the role described earlier, and in its Chief Data & Analytics Office, where a Hyderabad team "embeds with lines of business to deliver mission-critical, production-grade data solutions on Databricks and Snowflake." For the bank's in-house AI platform, see LLM Suite.

Two things set the bank seat apart from the vendor's:

• The customer is another department of the same firm, so travel is light. Morgan Stanley's posting listed none.
• The lessons stay in-house. A vendor engineer sends them back to a product team; JPMorgan's private-bank posting asks its engineer to feed "reusable patterns, skills, and learnings back into the team's platform so each engagement compounds."

The buy-side seat

Hedge funds and trading firms tend to be customers here rather than employers of the title. Point72, a hedge fund, and Jane Street, a trading firm, hire machine learning engineers under that title for their own teams, while the forward deployed engineers who work on a fund's problems come from the vendors selling to it.

The buyout firm is the exception worth knowing. Vista, whose posting set the month-six bar above, placed the role in its Product & Technology Practice and called it "the Palantir FDE model, applied to the largest enterprise software portfolio in the world." It also brings in other firms' engineers: under an April 2026 partnership, the two firms said Google Cloud would "allocate forward-deployed engineers (FDEs) to work side-by-side with Vista's portfolio companies and Vista's Value Creation Team."

What skills does a forward deployed engineer need?

Read side by side, the postings ask for the same four skill sets whichever the seat. The first three are an engineer's; the fourth is where a finance background earns its place.

Production code in more than one language

Python appears in almost every posting, usually with a second language set by the customer's stack. Morgan Stanley's research role wanted "Strong Core Java, SpringBoot, Spring AI," Apache Kafka for streaming, and "Python proficiency with at least one framework (FastAPI, Flask, or Django)." Vista asked for comfort "ramping quickly across languages and stacks" such as Python, Java, Go, TypeScript and C#, depending on the company. Most also ask for cloud, containers and deployment pipelines.

Systems built on language models

The AI half of the job is building systems around existing models rather than creating new ones:

• retrieval, agents and the plumbing between them, which Morgan Stanley listed as "LLM integration, Agentic, RAG, MCP, A2A, Tool integration, Workflow automation"
• "Deep, hands-on experience with LLMs and AI agents as well as core machine learning concepts," in Citi's words
• evaluation frameworks that show a system works on the customer's own cases, which the labs' postings ask for by name

Discovery and translation

The postings also want someone who can find the problem before solving it. Citi's describes "the crucial bridge between business needs and technical solutions"; JPMorgan's asks for "eliciting needs, framing problems, demoing, influencing, and building trust across technical and business audiences." Even where a separate lead owns the statement of work, as at the lab above, the engineer still runs discovery with the customer's staff.

The finance layer

This is the part an engineer from another industry has to learn:

• Morgan Stanley lists "Finance business understanding" as a requirement.
• Citi prefers "Knowledge of a specific business domain such as payments, finance and operations in financial services firms."
• JPMorgan's private-bank role prefers experience in "wealth, private banking, or asset management."
• The lab posting written for finance asked for experience of "the financial services industry and/or investment lifecycle."

Underneath all of them sits the constraint Ganesh described: numbers that have to be right, with a trail that shows why. A candidate who has reconciled a model to a filing, or traced a figure through three systems to find where it changed, already has the habit these deployments are built around.

How do you become a forward deployed engineer?

Most of the seats in the table want several years of engineering first. There are three common routes in, and only one starts at graduation.

Straight from university

Palantir is the one employer in that table that hires graduates straight into the title. Its new-grad postings promise responsibility from the first day: "As a New Grad, you will be entrusted with responsibility on day 1. You will not be handed a ticket queue." They ask for an engineering degree, ideally in computer science, mathematics or a similar field, proficiency in at least one programming language, and "Thoughtful responses to our application questions."

Its "Year at Palantir" internship starts earlier still. It is "open to freshmen and sophomores, as well as juniors and seniors in college," and asks for "Hands-on experience applying AI (whether through coursework, personal projects, internships, or relevant extracurricular activities)."

Through a consultancy or a data platform

Deloitte's Databricks role shows the shape of this route. It wanted three years in software, data engineering, data science or analytics engineering, a year of building generative AI systems, and a year on Databricks itself, with half of working time on the road. Platform skills of that kind carry into the bank seat, where the same tools run.

From the finance side

Citi's AI business analyst version is the clearest opening for someone whose background is business analysis rather than pure engineering. It asked for "4+ years of relevant experience in business analysis, systems integration, or technical consulting," and the work mixes requirements with building: "developing custom scripts, integrations, or workflows to tailor AI/GenAI products to the unique needs of the business."

Whichever route, the evidence that carries is something you built that other people used, and a record of how you checked it.

What do forward deployed engineer interviews test?

The interview is where that evidence gets tested, and the labs publish how theirs work. OpenAI's interview guide describes a skills-based assessment whose formats "may include: pair coding interviews, take-home projects, technical tests," followed by "4–6 hours of final interviews with 4–6 people over 1–2 days." For engineering interviews it looks for "well-designed solutions to the challenge, high-quality code, optimal performance, and good test coverage," and it says it evaluates "strong communication and collaboration skills."

Rules on AI tools differ by stage. OpenAI says some formats "intentionally allow them, while others are designed to assess your independent problem-solving without AI tools." Anthropic's guidance for candidates asks applicants to write a first draft themselves and then use Claude to refine it, but wants take-home assessments done without it "unless we indicate otherwise," and live interviews with "no AI assistance unless we indicate otherwise."

Beyond the format, the requirement lists in the postings are the best guide to the questions: expect the four skill sets above, including the finance layer wherever a bank lists it as a requirement.

The interview readiness check tests the AI tools, roles and bank systems a finance interviewer expects you to know, and the wider library of guides covers what the banks have built.

How the forward deployed engineer role is changing

The employers changed shape in 2026. The labs set up separate companies to do the deploying, and finance firms put money behind them.

The labs build deployment companies

In May 2026, OpenAI launched the OpenAI Deployment Company, majority-owned and controlled by OpenAI, which it said would start with "more than $4 billion of initial investment." Its purpose is to "embed engineers specialized in frontier AI deployment, known as Forward Deployed Engineers, or FDEs, into organizations working on complex problems in demanding environments." OpenAI also agreed to buy an applied AI consulting firm that would bring "approximately 150 experienced Forward Deployed Engineers and Deployment Specialists."

The partnership behind it is led by TPG, with Advent, Bain Capital and Brookfield as co-lead founding partners and Goldman Sachs among the founding partners; McKinsey and Bain & Company are among the investors.

In July 2026, Anthropic, Blackstone and Hellman & Friedman introduced Ode with Anthropic, a standalone AI services firm whose investors include Goldman Sachs, General Atlantic, Apollo and Sequoia. Garvan Doyle, Anthropic's head of forward deployed engineering for the Americas, said mid-size companies moving AI into their operations "need partners with real implementation depth and a clear understanding of how their businesses actually work."

For a finance candidate the pattern is worth noticing: some of the same banks and buyout firms that hire analysts now hold stakes in the firms that employ forward deployed engineers. The labs are also building industry teams in-house: OpenAI's board in September 2026 carried healthcare and legal versions of the job, and Anthropic was hiring a manager of technical deployment for financial services.

Banks build their own bench

The bank postings point the other way, toward owning the skill rather than renting it: one asks its engineers to train the experts who follow them, another to feed each engagement's lessons back into its own platform.

In May 2026, CIO quoted a prediction from Gartner senior director analyst Alex Coqueiro that by 2028 "70% of enterprises will be forced to abandon agentic AI solutions from FDE-led engagements because of high vendor costs and lack of internal skills to evolve them independently."

The bottom line

A forward deployed engineer in finance is still, most of the time, someone else's engineer sitting inside a bank or a fund: a lab's, a platform's or a consultancy's. By 2026 there was also a second seat inside the banks themselves, a buyout firm recruiting its own, and deployment companies backed by the same institutions the engineers are sent into.

The core of the work is the same in every seat. Learn how the institution really runs, ship something it depends on, prove the numbers, and leave it able to run without you. Engineering skill is the entry ticket; the habit of proving every number is where someone who already knows finance starts ahead.