Senior Data Scientist

Description

A team from the company is building a predictive modeling platform that runs on a live ad exchange handling hundreds of millions of auction requests each day.

The platform scores ad supply in real time, predicts high-performing audience contexts, builds look-alike audiences from small seed datasets, and optimizes campaigns against several business goals and operational limits.

The work covers censored bid-landscape modeling, sparse and delayed conversion attribution, calibration, counterfactual evaluation, and constrained optimization.

The customer operates supply-side infrastructure in programmatic advertising and is investing in predictive decisioning to improve performance, audience targeting, and campaign optimization.

The solution is built for large-scale production and is developed closely with the customer's internal data science team.

Responsibilities

• Build censored bid-landscape models that estimate clearing-price distributions from partly observed auction data.
• Develop real-time win probability models that respond to changes in bid pricing.
• Design and implement hierarchical lift estimation models and selection strategies based on confidence bounds.
• Build conversion propensity models that work with sparse, delayed, and aggregate-only labels.
• Create look-alike audience models using positive-unlabeled learning and embedding-based nearest-neighbor methods.
• Apply calibration per advertiser while tracking ranking and calibration quality on your own.
• Set up robust offline evaluation using inverse-propensity scoring, doubly-robust estimators, and importance reweighting.
• Define exploration strategies and propensity logging so that downstream correction and evaluation stay reliable.
• Develop constrained optimization for campaign objectives, pricing limits, and volume targets.
• Run data diagnostics, capability assessments, and turn findings into evidence-based model recommendations.
• Work with the customer team during post-launch tuning and performance validation.
• Write technical documentation and knowledge transfer material for the customer's data science team.
• Take part in architecture discussions and help shape scalable ML platform decisions.

Requirements

• You have 5+ years in machine learning or data science, including production models judged against business KPIs.
• You are strong in Python, including numpy, pandas, and scikit-learn.
• You write strong SQL and have worked with large datasets.
• You have hands-on depth with XGBoost, LightGBM, or CatBoost.
• You understand regularization, calibration methods, and categorical feature handling well.
• You have a solid grasp of probability, statistics, confidence intervals, and statistical power analysis.
• You have done feature engineering on structured and behavioral data.
• You have hands-on experience with Spark or PySpark.
• You have practical knowledge of experimentation frameworks and A/B testing.
• You have used advanced validation methods such as temporal splits, leakage detection, drift analysis, and slice-based metrics.
• You can apply explainability techniques such as SHAP and permutation importance.
• You speak English at Upper-Intermediate level or higher.
• Plus: experience in AdTech modeling, including CTR/CVR prediction, bid-landscape modeling, audience segmentation, and RTB mechanics.
• Plus: experience with sparse, delayed, or censored labels.
• Plus: knowledge of attribution modeling, survival analysis, and positive-unlabeled learning.
• Plus: practical experience with counterfactual and off-policy evaluation.
• Plus: understanding of isotonic regression and Platt scaling.
• Plus: experience with hierarchical, empirical-Bayes, or partial-pooling models.
• Plus: knowledge of constrained or multi-objective optimization.
• Plus: experience with uplift modeling and causal inference.
• Plus: experience with Vertex AI or similar managed ML training environments.
• Plus: publications, competitive modeling results, or open-source contributions in ML or AdTech.
• You bring strong analytical and problem-solving skills and do well in a highly data-driven setting.
• You communicate well and manage stakeholders effectively.
• You can explain complex modeling decisions to both technical and non-technical audiences.
• You take ownership proactively, and you pay close attention to detail and scientific rigor in experiments and evaluation.

Benefits

• You will work on technically challenging products alongside experienced engineers and data scientists.
• Your work will have a direct impact on large-scale production systems.