Picture two applicants for the same role. One has a GitHub page full of tutorial apps. The other has one small app that people in her neighbourhood use to book football pitches, a short write-up of what she learned from them, and pull requests that show how she checked every change an AI agent drafted. Read what product teams write in their job posts, and it is the second applicant they describe.

This guide draws on our own count of public company job boards. On 26 Oct 2026 we found 459 open Product Engineer roles at 301 companies, and 74% of them build with LLMs and AI agents. For the details below, we read the same boards again on 28 Oct 2026.

What does a product engineer actually do?

A product engineer owns a product, not just a piece of code. PostHog, which hires for the title, describes a product engineer in its handbook as a software developer who builds products for real users and writes full-stack code. In PostHog's framing, product engineers own the product and whether it succeeds, while software engineers own their code and leave product decisions to product managers.

In job posts, that ownership is concrete. incident.io's product engineers spend real time in the code and on customer calls, and own the work from design to the follow-up with the customer. At Ashby, engineers research competitors, write specs and make wireframes. PostHog adds a weekly support shift, some on-call time (being the person paged when production breaks) and writing the docs.

Hiring posts mention product engineers more often than they did. 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. Y Combinator startups use the title too. Of the 234 Y Combinator startups hiring engineers on YC's public jobs pages on 26 Oct 2026, 31 (13%) had a Product Engineer opening. See What is a product engineer? and our comparison of product, forward deployed and AI engineers.

How much experience do product engineer jobs ask for?

When they give a number, usually five years or fewer. Of the postings we read on 28 Oct 2026, 179 state a number of years. Of those, 134 (75%) ask for five or fewer, and the median is 5. About a quarter ask for two or fewer, and about a quarter for six or more.

Years of experience asked in Product Engineer postings 75% ask 0–5 years 2 11 35 24 15 47 17 14 9 0 5 0 1 2 3 4 5 6 7 8 9 10
Years asked (bottom) and number of postings (above the bars), among the 179 that state years. Green: the median. Shaded: 0 to 5 years, 75%. Source: our own count of public company job boards, 28 Oct 2026.

The bottom rung is the hard part. Only about 1 in 20 roles we read on 28 Oct was junior, graduate or intern level, while about 4 in 10 titles said senior, staff, principal or lead. Ashby says more than 90% of its engineering team is senior or above, so it has little room to mentor. SignalFire, a venture capital firm, finds new-graduate and entry-level hiring about 65% below its 2019 level at the 12 big tech companies it tracks, and about 76% below at early-stage startups. More in our look at tech hiring trends in 2026.

But the door is open. Granola says it plainly: "We don't set a minimum number of years." incident.io advertises a Product Engineer Graduate role in London, plus internships, and treats graduates like full-time engineers. On YC's jobs pages on 26 Oct 2026, about 3 in 10 Product Engineer openings said new grads are OK, though the sample is small. Where years don't count, proof does.

What skills do you need to become a product engineer?

The postings asking for 1 to 5 years stress users, teamwork, communication and ownership more than any specific tool.

The posting talks aboutShare
Users and customers88%
Performance and scale77%
Working across teams71%
Server-side work and APIs71%
Communication60%
Ownership58%
AI agents54%
Shipping and iterating50%

Share of the 132 postings asking for 1 to 5 years whose text mentions each theme. "Users and customers" means the posting talks about them, not that it requires customer calls. Source: our own count of public company job boards, 28 Oct 2026.

A smaller sample agrees. Of 113 postings we read by hand in October, two thirds asked engineers to talk to customers or users directly, and about 3 in 10 let engineers decide what to build. So build these habits:

  • Turn what users say into a small change you can ship this week.
  • Write things down: specs, pull requests, docs. 46% of the roles we read on 28 Oct were remote , and remote teams run on writing.
  • Measure: name the number your change should move, then check it after you ship.
  • Own what breaks: tests, logs, support.

What tech stack should a product engineer learn?

If you are starting from scratch, learn what the postings name most: TypeScript and React, then Python, PostgreSQL, AWS and Node.js.

Technologies named in Product Engineer postings TypeScript React Python PostgreSQL AWS Node.js Kubernetes Next.js Go 55% 46% 36% 25% 25% 21% 17% 15% 12%
Share of Product Engineer postings that name each technology. Source: our own count of public company job boards, 28 Oct 2026.

Linear works across the full stack in TypeScript end to end, with React, Node, GraphQL and PostgreSQL. Granola lists TypeScript, React, Node.js, Postgres and AWS. incident.io says its own stack of Go, TypeScript and React is not a requirement. So pick one full-stack set and go deep enough to ship and run a real app on a cloud provider.

How do product engineers work with AI coding agents?

Two skills hide behind "AI" in these posts. The first is building products with AI inside. Linear asks engineers to build AI-powered features, and its stack includes an agentic workflow system in production. PostHog requires applicants to have "built things agents actually use": an API an agent can drive, an MCP server or evals. Side projects count. MCP (Model Context Protocol) is a standard way for agents to call tools and data; evals are tests that score an AI system's output.

The second is building with coding agents. About 4 in 10 postings we read on 28 Oct mention AI coding tools or AI-assisted development. Ashby's engineers use AI to triage customer issues, prototype, write code and review it, with an unlimited token budget they are not measured on.

The skill that matters most is review. In Stack Overflow's 2026 developer survey, 48% of respondents trust AI when they can easily verify its output, and only 7% trust it with important decisions. Your job is to make the agent's work easy to verify, then verify it.

The review loop for working with a coding agent Spec Draft Check Review Ship Learn You own the loop
The review loop. Green steps are yours. Blue steps are done by the coding agent and your automated checks.
  1. Spec. Write what the change is for and how you will know it worked. Keep a short file of project rules in the repo.
  2. Draft. The agent writes the first version on a branch, in small steps.
  3. Check. Types, linter, tests and a CI build run before you look. This is the harness: guides before the agent writes, sensors after (see what harness engineering is).
  4. Review. Read the diff as if a new teammate wrote it: empty input, double clicks, needless packages, changes to sign-in, personal data in logs. If you can't explain a line, it doesn't merge.
  5. Ship. Merge small, behind a feature flag if it is risky, and say in the pull request how you used AI.
  6. Learn. Watch the metric and the support inbox, then write the next spec.

What should a product engineer portfolio show?

Product sense is hard to claim and easy to show. Build one or two real products, not ten tutorials, and write a one-page case study for each:

  • The user and the problem, and how you found out.
  • What you chose not to build. Saying no is product sense you can see.
  • A number before and after. Small is fine; made up is not.
  • One real pull request: the spec, the agent's draft, what your checks caught, what you changed.
  • What broke, how you fixed it, and the test that now guards it.
  • Something an agent can use: a small API or MCP server with a few evals, the kind of side project PostHog counts.

Be ready to show the work live. Few postings in our 28 Oct read describe their interview process , but some that do test real work. Ashby pairs with candidates for an hour in its actual codebase, then holds non-coding rounds on product thinking, technical design and infrastructure. It says "There are no leetcode or whiteboard exercises", meaning no timed algorithm puzzles. Linear's blog says its interviews end with a paid work trial of 2 to 5 days on a real project.

What is a good 90-day plan to become a product engineer?

This plan assumes you can already write basic code.

Days 1–30

Build the base

  • Find one real user with a real problem: a shop, a club, a family business.
  • Build a small full-stack app for them in TypeScript, React and Node.js on PostgreSQL, and deploy it to a cloud provider.
  • Write tests from the first week, with CI running them on every push.
  • Keep a short daily log in English of what you built and why.

Days 31–60

Ship with an agent

  • Work with a coding agent inside the review loop above.
  • Ship a small change every week, with basic analytics to see whether it gets used.
  • Talk to your user after each change.
  • Add one AI feature, such as search or a summary, with evals, including a prompt-injection test (input that tries to make it ignore its instructions).

Days 61–90

Prove your work

  • Give your product a small API or MCP server, with docs written for a machine.
  • Write the case study and record a short demo.
  • Get a fix merged into an open-source project you use.
  • Practise: pair on your own code, debug unfamiliar code against the clock, and explain a product decision out loud.
The takeaway

Most postings that state years ask for five or fewer. Years aside, build proof: a product with users, and agent-written code you checked and can explain. Ninety days is enough to start.

Frequently asked questions

How long does it take to become a product engineer?

There is no fixed time. Of the postings that state years in our 28 Oct read, 75% ask for five or fewer , and some teams set no minimum. If you can already code, 90 focused days can build the proof: a product people use, a case study and pull requests you reviewed.

Do I need a computer science degree to become a product engineer?

Not always. In our 28 Oct read, about 4 in 10 postings asking for 1 to 5 years mention a degree. Graduate schemes such as incident.io's ask you to be studying computer science or a related subject. Without a degree, shipped work is your strongest evidence.

Is a product engineer the same as a full-stack engineer?

The code is similar; the ownership is not. On 26 Oct 2026, Y Combinator filed 35 of the 43 Product Engineer openings on its public jobs pages under Engineering / Full stack , but a product engineer also owns whether the product works for its users.

Do product engineers have to use AI coding agents?

Many do. About 4 in 10 postings in our 28 Oct read mention AI coding tools , and 74% of the roles in our 26 Oct 2026 count build with LLMs and AI agents. What sets you apart is reviewing what the agent writes.

How is a product engineer different from a forward deployed engineer or an AI engineer?

A product engineer owns a product and its users. LinkedIn describes the forward deployed engineer as a role that helps organizations put AI solutions in place , and says it is now the third most common AI job in US job postings. Our comparison covers all three, including the AI engineer.

Sources

How we counted. About 360 public company job boards (Ashby, Greenhouse, Lever, Workable, Recruitee); software roles titled Product Engineer, posted in the last 180 days, hardware left out, one role in many cities counted once. Headline figures: 26 Oct 2026. Breakdowns: a second read on 28 Oct 2026. YC's public jobs pages show its most active companies, not all of YC. Postings read on 29 Oct 2026.

Product EngineeringAI Coding AgentsCode ReviewPortfolioLearning Path

Want to learn this with a cohort?

Nilerobot Academy is a paid Product Engineer training programme: build full-stack products with AI agents and learn to review what they write, on a foundations track (final-year students and fresh graduates) or a professional track (working developers). The academy is a separate training programme; it isn't part of hiring.

See the academy →