Salary: Up to £220,000 base + equity + bonus
The role
Most recommender problems optimise a feed. This one optimises a market: tens of millions of users earning real rewards for engaging with advertisers’ products, where every ranking decision carries a price, an incentive and a margin. Get the models right and revenue moves that week; get them wrong and the economics break. It’s one of the few consumer‑scale ML problems where ranking, pricing and incentive design collide in a single system — inside a company that is profitable, growing fast, and funds all of it without a VC in sight.
The work
- Own real‑time recommendation, ranking and pricing models at consumer scale — engagement and conversion are the direct revenue levers, not proxy metrics.
- Design reward and user‑incentive systems that balance what motivates users against sustainable ROAS, through uplift modelling and pricing optimisation.
- Drive experimentation where the money is: causal impact estimation, MMM and Multi‑Touch Attribution that change decisions, not decorate dashboards.
- Take zero‑to‑one bets from blank slate to production data product — initial data structures through final deployment, owned by you.
- Shape the stack and system architecture with engineering, and define the gold standard a 20‑person hub will inherit.
What you bring
- Recommender systems or ranking shipped in production at scale, owned end to end — the hard requirement, not learned on the job.
- A Staff‑level operating record: you set technical direction, create structure in ambiguity, and pull stakeholders along — while still writing the code yourself.
- Fluency across Python, SQL and the modern data stack (DBT, Airflow, Kubernetes, Terraform) on cloud, GCP preferred — ideally from gaming, adtech or user‑incentive systems.
- Eligibility to work in the UK and willingness to be on‑site in London.
As part of the recruitment process at Dex, we process your personal data in accordance with our Privacy Notice for Job Applicants. This notice explains how and why your data is collected and used, and how you can contact us if you have any concerns.
#J-18808-Ljbffr…