описание
BidMachine, Appodeal's in-app ad exchange, runs on machine learning. We are looking for an Exchange Product ML Lead who treats that ML as a product and owns it end to end: what gets built next, how it's tested, whether it ramps or stops, and the gross profit that follows. This is the DRI seat for Exchange ML: you lead a squad of five to six people and stay hands-on enough to read the code, challenge a model and run an experiment. Retraining, feature pipelines and monitoring are already automated; the work is deciding what to build next, designing the test that proves it, and making the call to ramp or stop.
задачи
- Set the product vision for ML in the exchange and turn it into a roadmap and PRDs the squad can build from.
- Own the gross-profit result of the squad's ML work.
- Run a weekly cadence of new models, features and embeddings on a unified dataset shared across business units, and decide which iterations go to test.
- Design A/B tests, read them out under agreed decision rules, and decide whether each iteration ramps, changes or stops.
- Weigh exchange margin against supply-path efficiency and what publishers earn.
- Review models and calibration before they get more traffic.
- Dig into bid and auction data, writing SQL and model code where needed.
- Debug cases where production behaves differently from offline results.
- Make the squad faster: shorter training time, a more efficient pipeline, fewer steps between a good result and full rollout.
- Lead the squad end to end and redirect its focus when the data says the current direction isn't paying off.
требования
- Owned an ML-driven AdTech product end to end: set the roadmap, wrote the PRDs, and can name the gross-profit or margin result it delivered.
- Designed A/B tests on ML changes, read them out and made the call to ship or stop.
- Hands-on: writes SQL and reads model code; can review a model and its calibration before a ramp and explain production vs offline differences.
- Knows auctions, exchange margins and supply-path optimization, and how a margin gain can cost publishers revenue.
- Has led a squad or team to a result, set its focus and changed it when the data said so.
- AI-native: uses AI tools daily, understands how they work, can describe workflows improved, results and how output was checked.
- Strong data skills on large bid and auction datasets; can explain findings to someone outside ML.
- Based in Barcelona or ready to relocate; 4 days a week in the office.
- Будет плюсом: Experience inside an ad exchange or SSP; Personally built exchange models (bidding, floor pricing, traffic shaping); Shortened model training time or automated model rollouts
условия
- With a powerful product, strong market positioning, and a team that strives for excellence, BidMachine offers a unique opportunity to grow, innovate, and make a tangible impact in the programmatic advertising space.