Summary
What you’ll impact
The Senior Data & Machine Learning Engineer will independently own the end-to‑end matching stack, building models, serving paths, and underlying platforms for a programmatic ad network. This high‑autonomy role involves designing experiments, developing infrastructure, and driving 0‑to‑1 initiatives from prototype to production.
Responsibilities
What you'll do
- Retrieval, ranking, and re-ranking for ad-to-recipient matching.
- The embedding and vector-search infrastructure powering real-time ad selection inside publisher newsletters.
- Multi-objective optimization in a single ranking stack: CTR lift, advertiser ROAS, publisher revenue, and subscriber relevance.
- The feature store and the serving path: offline training, online inference, and the parity between them. Pipeline health is your problem, not someone else's.
- LLMs and foundation models applied to subscriber tagging, content understanding, and user modeling.
- Data pipelines that turn raw event data into reliable, queryable, experiment-ready datasets.
- The warehouse and transformation layer.
- Experimentation end-to-end: find the high-value questions, design statistically sound A/B and multivariate tests, instrument collection, analyze, and recommend.
- Dashboards and automated scorecards giving the team and leadership real-time visibility into performance, CTR lift, tag coverage, and other key metrics.
- 0 to 1 initiatives from prototype to production.
- Clean data contracts and well-designed schemas, in partnership with engineering.
Requirements
What you’ll bring
- 5+ years of production data or ML engineering experience where you owned outcomes, not just deliverables.
- Production recommendation or matching systems: candidate generation, ranking models, and embedding-based retrieval. You've shipped a system that does all three layers, not just one.
- Vector database experience in production, not a prototype.
- Strong SQL and data modeling. You're fluent in Postgres and comfortable across columnar analytical databases.
- Hands-on dbt experience. You've built and maintained transformation layers, written tests, and used dbt (or similar) as a core part of a production stack.
- Data pipeline engineering chops. You can build reliable ETL/ELT workflows, not just consume their output.
- LLMs applied to structured prediction. You're comfortable using foundation models for tagging, enrichment, and classification.
- Online learning and feedback-loop systems, where the next training cycle depends on yesterday's serving logs.
- Experiment design rigor. You understand statistical significance, sample sizing, and the difference between a compelling story and a valid conclusion.
- Independent, self-directed working style. You don't wait for someone to define the question. You find it, scope it, and go after it.
- Comfort with ambiguity and imperfect data. You know how to work with what's available while building toward what's ideal.