Picture one request: a logistics company wants its staff to ask questions of thousands of shipping documents and get answers they can trust.
A product engineer at a software company asks whether this should be a feature for every customer, ships a small version and watches who uses it.
A forward deployed engineer from the software vendor the firm just signed with goes to the customer, connects its data and ships the tool inside its systems.
An AI engineer owns the middle: which documents the model sees, how its answers are tested, and what happens when a document tries to give the model orders.
Same request, three different jobs.
What is the difference between a product engineer, a forward deployed engineer and an AI engineer?
The product engineer is anchored to the product. PostHog's plainest definition is “a software engineer building products”. Product engineers own the product and whether it succeeds; software engineers own their code. No one company can be credited with the title; the earliest software use we could verify is at Intercom in early 2012. More in What is a product engineer?
The forward deployed engineer is anchored to one customer. Palantir, which says it pioneered the role, contrasts a software engineer who builds one capability for many customers with a forward deployed engineer who brings many capabilities to one customer and is judged by that customer's results. A 2019 Palantir post shows the title in use there by about 2012. More in What is a forward deployed engineer?
The AI engineer is anchored to the model. Chip Huyen, author of AI Engineering, defines the work as building applications on ready-made foundation models, large general-purpose models trained by someone else: less model training, more prompts, context and light fine-tuning (a little extra training on your own examples). An essay by swyx on 30 June 2023 made the title widely known. More in What is an AI engineer?
How do the three roles compare side by side?
| Question | Product engineer | Forward deployed engineer | AI engineer |
|---|---|---|---|
| Focus | One product and whether its users get value | One customer's goal | What the model sees and how its answers are tested |
| Who they work with | Designers, product managers, researchers, data analysts and users | The customer's people up to executives; sales and product; at some firms a deployment strategist, who scopes problems and trains users | Builds on models made by researchers and ML engineers, through APIs |
| Typical day | At incident.io, part code, part customer calls; at PostHog, a support shift and some on-call (being paged when production breaks) | At Palantir: scoping with the customer, deploying, going live with users, training their IT team | Prompts, context, light fine-tuning, evaluations |
| Core skills | TypeScript and React most named, then Python; users, teamwork, ownership | Python, then TypeScript or JavaScript, APIs and integrations; communication | Agents, retrieval, prompts, guardrails; Python first |
| Years employers ask | Median 5; 75% of ads that state years ask five or fewer | Median minimum 5; three in four ask 3 to 8 | No ad count of ours; on LinkedIn, US AI engineers had a median 3.7 years' prior experience |
| Where the work happens | In the company's own product team; 46% of roles remote | At the customer; many ads state travel of up to about 50% | Varies; on LinkedIn, US openings were 26.2% remote, 27.1% hybrid (posted Nov 2024 to Nov 2025) |
Each column uses different sources and counts, so read down a column, not across. Product engineer figures are from the roles we read on 28 Oct 2026.
How is a forward deployed engineer different from a product engineer?
These two get mixed up most: both write full-stack code and talk to customers. One scan of 294 company job boards on 12 Oct 2026, with the same keyword rules for both titles, shows the gap.
- Forward deployed engineer ads
- Product engineer ads
Forward deployed ads mention customer-facing work (with customers, stakeholders or executives) more than twice as often, and on-site work or travel about four times as often. Product engineer ads lean to TypeScript or React and to design. Both had a median minimum of 5 years.
Is it sales? It depends. Salesforce files its forward deployed job under Sales; Figma says its role is “not a sales or services function”. Pre-sales work, such as trial builds that help a customer decide to buy, mostly belongs to solutions engineers: 74% of their ads in the same scan mention it, against 25% of forward deployed ads. Nor is it consulting, says Palantir: its engineers build lasting systems, not one-time analyses, and often add features to the core product.
Is an AI engineer just a machine learning engineer?
Not in the job ads. Our own count of 1,768 public company job boards on 29 Oct 2026 found 223 open AI Engineer roles at 136 companies and 220 Machine Learning Engineer roles at 129, and the two read like different jobs.
- AI engineer ads
- Machine learning engineer ads
Retrieval, or RAG, means handing the model the right documents along with the question. Evals are tests that score an AI system's answers; Huyen notes that open-ended answers make these models harder to evaluate. Guardrails are checks around the model; the OWASP list of LLM risks starts with prompt injection, where text slips the model new instructions.
The 2023 essay drew the border as an API line: ML researchers and engineers work on the models, AI engineers build on top through APIs, and people cross both ways. In swyx's words: “When it comes to shipping AI products, you want engineers, not researchers.” LinkedIn adds a caution: AI Engineer overtaking Machine Learning Engineer in US postings may partly reflect companies relabelling jobs once called machine learning engineering.
How many of these jobs are there, and are they growing?
Each count below was taken differently, on a different day, so together they show that all three jobs are real, not which is bigger.
- Product engineer. Our own count of about 360 public company job boards found 459 open Product Engineer roles at 301 companies on 26 Oct 2026, and 74% of them build with LLMs and AI agents. (LLMs are large language models.) In our count of Hacker News “Who is hiring?” posts, 7.6% mention Product Engineers (January to October 2026), up from 5.2% in 2025 and 3.9% in 2024.
- Forward deployed engineer. On 29 Oct 2026, Palantir's job board had 75 open ads with “Forward Deployed” in the title, Databricks' 114 and OpenAI's 32, counting every posting on each board. SignalFire, a venture capital firm, reports that Forward Deployed Engineers' share of engineering roles rose 30% since 2022; it counts people in roles, not job postings. LinkedIn calls it the third most common AI job in US job postings.
- AI engineer. LinkedIn's Jobs on the Rise lists, which rank titles by growth in members starting them, put AI engineer first in the US in 2025 and 2026; the 2026 list includes machine learning engineers in that title. LinkedIn also says AI Engineer has overtaken Machine Learning Engineer as the most common AI role on its site.
Two of the titles draw young hires: LinkedIn's Economic Graph report says Gen Z is more than two-thirds of US hires into Forward Deployed Engineer and AI Engineer roles. The wider market is tougher: Indeed's US software development postings were still about a fifth below their February 2020 level on 23 Oct 2026. More in our tech hiring trends for 2026.
Where do the three roles overlap?
- Product and customer: talking to users. In our hand-read sample of 113 Product Engineer postings (11 Oct 2026), two-thirds asked engineers to talk to customers or users. Figma asks its forward deployed engineers to turn each fix into a reusable path for future customers.
- Product and model: AI features in a product. PostHog now asks Product Engineer applicants to have built something AI agents use. AI engineer ads look like product jobs too: 31% mention TypeScript or Node.js, against 6% of machine learning ads.
- Customer and model: AI at a customer. LinkedIn describes the forward deployed engineer as helping organisations put AI in place, and about half of the forward deployed ads we read mention AI agents.
- The middle: owning the result. Most postings in our hand-read Product Engineer sample and most of the forward deployed ads we read ask engineers to own the work end to end, and the 2023 essay makes product data and evals the AI engineer's job.
AI coding tools also come up in about 4 in 10 Product Engineer postings (28 Oct 2026). Making their output trustworthy is a discipline of its own: harness engineering.
Which role should you choose?
Titles shift, so choose by the work you want to be judged on.
Who do you answer to? A product's users, for years: product engineering. One customer's results, with new problems on every account: forward deployed work. Whether an AI system's answers are right, safe and affordable: AI engineering.
How much travel suits you? At OpenAI, 23 of 25 forward deployed ads state travel of up to or about 50%. Product engineering is often remote: 46% of the roles we read on 28 Oct 2026.
What do you like to debug? User behaviour, other companies' systems, or answers that are nearly right. That last one is common: 66% of developers in Stack Overflow's 2025 survey named AI solutions that are “almost right, but not quite” as their biggest frustration.
Where can you start? Product engineering has entry points such as incident.io's graduate role. Forward deployed ads want a median minimum of 5 years, though Palantir posts internships. On LinkedIn, AI engineers most often came from software engineer, data scientist and full-stack roles.
Our view at Nilerobot: if you are unsure, start with product engineering, because owning a product is the habit the other two build on. Then read how to become a product engineer.
The three roles share a stack and a habit of owning results. What differs is the anchor: a product's users, one customer's goal, or the quality of what the model does. Pick the anchor you want; the title will follow.
Frequently asked questions
Is a forward deployed engineer a consultant?
Palantir, which says it pioneered the role, says no: consultants hand over a one-time analysis, while its forward deployed engineers build lasting systems and do far more engineering. SignalFire, a venture capital firm, describes them as acting like consultants who help customers adopt AI.
Is an AI engineer the same as a machine learning engineer?
They overlap, but the job ads differ. In our own count of public company job boards on 29 Oct 2026, AI engineer ads mentioned AI agents 79% of the time against 31% for machine learning ads, while machine learning ads mentioned PyTorch, TensorFlow or JAX 65% of the time against 13%.
Do product engineers need AI skills?
More and more. Our own count of about 360 public company job boards found 459 open Product Engineer roles at 301 companies on 26 Oct 2026, and 74% of them build with LLMs and AI agents.
Which role is easiest to get as a new graduate?
None is easy, but each has a way in. For example, incident.io advertises a Product Engineer graduate role, and Palantir posts forward deployed internships, though the forward deployed ads we read had a median minimum of 5 years. LinkedIn reports that Gen Z is more than two-thirds of US hires into Forward Deployed Engineer and AI Engineer roles.
Do forward deployed engineers travel a lot?
Often. On 29 Oct 2026, 23 of OpenAI's 25 forward deployed engineer ads stated travel of up to or about 50%. Palantir's ads range from 25% to 75% by team, and Databricks often states 20%.
Sources
- What is a product engineer, PostHog, 1 Dec 2022; Product engineer vs software engineer, PostHog, 8 Dec 2025.
- Intercom's Darragh Curran on scaling engineering, Intercom, 15 Sep 2016.
- Dev versus Delta, Palantir Blog, 8 Apr 2019; Who Wants to be a Delta?, 21 Oct 2021.
- The Rise of the AI Engineer, Latent Space (swyx), 30 Jun 2023.
- AI Engineering, Chip Huyen (O'Reilly, 2025), with its README and chapter summaries.
- OWASP Top 10 for LLM Applications 2025, OWASP, 17 Nov 2024.
- New LinkedIn research on AI hires, LinkedIn, 18 Aug 2026; The AI Talent Divide (PDF), LinkedIn Economic Graph, Aug 2026.
- Jobs on the Rise 2025 (US), LinkedIn News, 7 Jan 2025; Jobs on the Rise 2026 (US), 7 Jan 2026.
- State of Talent Report 2026, SignalFire, 22 Jun 2026.
- 2025 Developer Survey: AI, Stack Overflow, 2025.
- Indeed Job Postings Index, Indeed Hiring Lab, data to 23 Oct 2026.
- Job postings (not linked; postings change and close), read 29 Oct 2026: Intercom (posting now under the Fin brand), incident.io and PostHog (product engineering); OpenAI, Palantir, Databricks, Figma and Salesforce (forward deployed).
- Our own data (Nilerobot): counts of public company job boards (about 360 boards, 26 Oct 2026, breakdowns read 28 Oct; 294 boards, 12 Oct; 1,768 boards, 29 Oct; 147 forward deployed ads, 28 Oct; three companies' boards, 29 Oct), 113 hand-read postings (11 Oct) and Hacker News “Who is hiring?” posts (26 Oct).
Need engineers who own the outcome?
Nilerobot is a software studio in Egypt. We design, build and run your product, or our engineers join your team, managed and mentored by us.
See how we work →