About us
## What you will do
- Build the performance feedback loop. Ad metrics in, creative patterns out.
- Train models that predict creative win rate before spend.
- Design ranking systems for creative variants under budget constraints.
- Partner with the generative team. Your signals guide their generation.
- Build dashboards the media buying team actually uses daily.
- Run experiments. Hold-outs, uplift tests, counterfactual analysis.
- Ship to production. Not notebooks.
About your position
Every ad that runs through Coinis becomes a data point. You turn that data into an optimization loop that makes the next creative smarter. Creative scoring, audience matching, budget allocation, ROAS prediction. This is the flywheel that makes Coinis pull ahead of every creative-only tool.
Job requirements
## Must have
- 5+ years in ML, data science, or applied stats with a shipping track record.
- Strong Python. Pandas, scikit-learn, PyTorch or equivalent.
- Causal inference and experimentation design experience.
- Production ML experience. Serving, monitoring, retraining.
- Comfort with messy real-world ad performance data.
## Nice to have
- Meta Marketing API data pipelines.
- Contextual bandits or reinforcement learning in production.
- Feature stores. MLOps tooling.
- Multi-armed bandit budget allocation.
Qualifications
01
Education
MSc preferred in Statistics, ML, Computer Science, or related. BSc with strong experience also welcome.
02
Experience
5+ years
03
Skills
Python, PyTorch, scikit-learn, Causal Inference, Experimentation, MLOps, Meta Marketing API, Contextual Bandits, ROAS Modeling, Feature Stores
What we offer
## What you get
- Competitive salary in EUR.
- Equity in the SaaS spinout.
- Access to live Meta ad spend data across thousands of advertisers via our media buying operations.
- Rare real data. Not a sandbox.
- Remote-first with flexible hours.
- Compute and research budget.
Other information
You will work directly with our in-house media buying team. They run active daily Meta ad spend. Your models power the optimization they do today and the agent we build tomorrow.
