AI search researcher
Find out how ChatGPT, Perplexity, Gemini, Claude and Google's AI search, read and pick their sources, and turn it into what Echo measures and recommends.
- Mid to senior level
- €60,000 base
- Remote
- Full time
The role
About the role.
Every AI assistant decides which pages to read and which to cite, and none of them says how. They change month to month. Echo sees a lot of it: the searches an assistant runs behind a question, the pages it fetches, the ones it cites, and what happens after a website changes.
As our AI search researcher you study that behaviour. You design the probes, keep the test pages, notice when an engine shifts, and decide what it means for the teams using Echo. You also build the labelled ground truth Echo's own AI checks are scored against, so we know when they're wrong.
It's research that ends up in the product: in the actions Echo suggests, in what it measures, and in what Echo Lab publishes.
What you'll do.
- Study how each engine searches, retrieves and cites, and keep track of when that changes
- Run experiments on test pages: what gets read, what gets cited, and how long it takes
- Build and keep the human-labelled sets Echo's AI checks are scored against (sentiment, fact checks, page types)
- Turn what you find into the rules behind Echo's actions, with the team
- Write up findings for Echo Lab, from short notes to full articles
What we look for.
- A background in information retrieval, search ranking, recommender systems or evaluating language models, in research or industry
- You've built an evaluation set and know how it fails
- Curious enough to keep a lab notebook, and careful enough to say what a result doesn't show
- You use AI tools every day and know where they go wrong
- You pick your own questions and don't wait for a brief
Mid to senior level: you've worked on search, retrieval or model evaluation and have results to show. Tell us about something you found out about a system nobody documented.
What you get
A base, a performance part, and room to work your way.
€60,000 a year
The base salary, paid monthly.
A performance part
A performance part tied to the findings that ship in the product and the research Echo publishes, agreed with you before you start.
Fully remote
Remote, within about three hours of Central European Time. No office, no offsites, no mandatory anything.
Your own show
You plan your own weeks and pick your own tools. We judge the outcome, not the hours.
How the pay works in full, and the questions people ask about it: the careers page.
How it tends to go
The first months.
- Weeks 1 and 2: learn Echo as a user, read the answers, searches and citations it collects, and pick the question you'd ask first
- Month 1: test pages live and the first labelled set built, with how well Echo's checks score against it
- Month 3: a first finding in the product's actions, and a first Echo Lab note written by you
In short
How we work here.
Remote by design
There's no office to come to, and there won't be one. You work from wherever you do your best work, in the hours that suit the customers you're talking to. We hire people who like it that way.
You run your own show
Nobody hands you a calendar. You decide how to find customers, how to run your week and which tools to use, as long as it adds up to new customers. Think of it as running your own business inside ours, with the product, the brand and the leads already there.
Paid for outcomes
We'd rather pay you a lot for results than a little for attendance. The base salary is deliberately modest; the share of revenue isn't, and it has no cap. Someone who does well here earns two to three times their base.
Light on rituals
No offsites, no mandatory socials, no all-hands for the sake of it. A short written update, a call when a call helps, and colleagues who respect your work. If you make friends here, great; it's just not part of the job.
Read this before you apply
The interview is a study design.
We don't do coding tests or whiteboard rounds. We'd rather see how you'd answer a real question:
- Apply below. A person reads it and gets back to you within a week.
- We send you an anonymised slice of Echo's data and a real question we want answered, for example how stable an AI answer is, or whether a change on a website moved it.
- Write a one-page design: the method, what would prove it wrong, and its limits. You don't have to run it.
- We talk it through on a call, with the people you'd work with, then you get a decision within a week.
If you'd rather be judged on how you think than on how you perform in an interview, this is for you.
Apply
A few lines are enough. Honest ones are better.
No CV upload, no cover letter. Your LinkedIn profile tells us where you've been; the box below tells us whether this is the right place for you.
- How you match this level. Mid to senior level: you've worked on search, retrieval or model evaluation and have results to show. Tell us about something you found out about a system nobody documented.
- How you'd work when nobody sets your week: how you'd organise yourself and decide what to do next.
- One piece of work you're proud of (a paper, a method, an evaluation, an experiment), with a link, and what it changed.