We're hiring in Tallinn. What the work is like.

Most job posts describe a role. This one mostly describes a loop, because everyone here works on some part of it.

Zobi's agents sell for commerce brands every day. Each conversation comes back labelled with what the customer did: replied, bought, went quiet, reordered. That trains the next version of the model, the next version sells a bit better, better selling brings in more conversations, and more conversations means more data. It's running while you read this.

What the days are like

Your work ships the week you start. There's no six-month roadmap between you and a customer. If you improve how the model picks a reply, real people get better messages that week and you can watch the reply rate move. If you break something you'll see that too. It's the fastest feedback most of us have had on our work.

You work on the real thing. The dataset isn't a benchmark somebody downloaded. It's over a million sales conversations with real outcomes, growing daily, and it doesn't exist anywhere else. If you've wanted to do research on the kind of data only production produces, this is it.

You're not one of two hundred. We're a small group of operators and researchers with no layer between you and the decisions. You'll argue about the reward function with whoever wrote it, which may well be you.

The roles

Machine Learning Engineer

Own the training loop, from outcome labels to deployed policy. How do we turn a million labelled threads into a model that sells better than the last one, and how do we prove it did with live holdouts? This is the centre of the company.

Senior Software Engineer

Build the systems a million conversations run through: sending, routing, reliability, and keeping the loop turning under real load across every messaging channel a customer might answer on.

Data Engineer

Turn raw threads into the cleanest sales dataset in the world. Labelling, lineage, quality, and the pipes that make sure what the model learns from is true.

Growth Operator

Put the model in front of the next hundred brands. Every brand you bring on means more conversations, more data and a better model, so this role feeds the loop as directly as the engineering ones do.

And the rest

We're also hiring a Research Scientist, an Infrastructure Engineer, a Customer Success Manager, Legal Counsel, a Trust & Compliance Lead and an Operations Manager. The full list, with what each one is for, is on the application page. The four above are the ones we're most impatient about; the others matter just as much once the loop is bigger than a handful of people can run.

Who we're looking for

Two things, and we mean both. First, people who are genuinely excellent at their craft: the ones other strong people go to when a problem is hard. Second, people with real drive. Not busyness, but the habit of taking a thing that matters, owning it end to end, and getting it done well and quickly, without waiting to be asked. The best people we've worked with have both, and it's a rarer combination than it sounds. Talent that doesn't ship is theory; effort without judgement is noise.

In practice that looks like: you'd rather find out than speculate. You've shipped something, watched the number, and changed your mind because of what you saw. You're comfortable being the person responsible. And you get more done in a week than most people expect, because you care about the outcome and you're good enough that caring translates into results.

If that's you, the specific stack matters much less than you'd think. If it isn't, no stack will make up for it. We're backed by Seedcamp and Carles Reina, we're in Tallinn, and we're small enough that this was written by the people you'd be working with.

Don't see your role? Write to us anyway: business@zobi.com.