Описание
Andersen is hiring a Staff Data Scientist / Machine Learning Engineer for a project developing an AI-powered advertising platform to predict customer lifetime value and optimize campaign performance. The project is focused on developing and enhancing an AI-powered advertising optimization platform that predicts customer lifetime value using machine learning models. It integrates predictive value signals with major advertising platforms to improve value-based bidding, optimize marketing budgets, and maximize campaign performance through accurate data calibration and automated signal delivery.
Задачи
•Building and maintaining predictive Lifetime Value (pLTV) models using XGBoost, CatBoost, and BG/NBD within a tiered scoring architecture, routing users into appropriate model tiers based on data availability and observability rather than applying a one-size-fits-all regression approach.
•Designing, implementing, and maintaining model calibration layers to transform raw predictions into platform-ready values, incorporating haircut scalars, linear, logarithmic, and rank-based normalization, and asymmetric loss functions that penalize overestimating low-value customers more heavily than underestimating high-value customers.
•Owning feature engineering processes in BigQuery, developing and maintaining Gold-layer feature tables from cleaned Silver-layer data, including RFM metrics, behavioral embeddings, category affinity, and purchase velocity.
•Ensuring feature quality and reliability by evaluating data completeness, monitoring missing-value rates, and excluding features with high missing rates among new users rather than relying on imputation.
•Monitoring model drift and distribution shifts, implementing residual logging, comparing pre- and post-signal deployment cohorts, and identifying causal contamination in retraining datasets resulting from the model's influence on user acquisition.
•Maintaining and improving batch Union-Find identity resolution algorithms in BigQuery, connecting GA4 behavioral identifiers with Shopify transactional identifiers and optimizing identity match rates to improve signal quality.
•Developing and maintaining production-grade Python applications running on Google Cloud Run Jobs, orchestrating workflows through Apache Airflow, and ensuring reliable delivery of ML-generated signals to advertising platforms.
•Implementing and maintaining automated testing, CI/CD pipelines, idempotent processing, and dry-run safety mechanisms to ensure production stability, prevent costly errors, and protect client advertising performance.
Требования
•Hands-on experience in Machine Learning Engineering using Python, including developing, deploying, and maintaining production-grade ML models, for 5+ years.
•Proven experience with gradient boosting algorithms (XGBoost, CatBoost, or LightGBM), including hyperparameter tuning, feature importance analysis, and advanced model evaluation.
•Strong proficiency in SQL and BigQuery, including complex feature engineering queries, scheduled data transformations, and data validation.
•Hands-on experience deploying and maintaining ML workloads on Google Cloud Platform (GCP), including Cloud Run, Airflow/Composer, Vertex AI, and Pub/Sub.
•Solid understanding of model calibration, business-oriented loss functions, and the impact of prediction errors on downstream optimization systems.
•Practical experience integrating ML model outputs with external systems through APIs, such as advertising platforms, recommendation engines, pricing systems, or bidding systems.
•Level of English – from Upper-Intermediate and above.
Будет плюсом
•Experience in e-commerce or Direct-to-Consumer (DTC) environments, particularly customer lifetime value (LTV) prediction, purchase forecasting, churn modeling, or customer segmentation.
•Familiarity with probabilistic customer models (BG/NBD, Gamma-Gamma, or similar frameworks) or strong foundations in Bayesian statistics.
•Experience with advertising platform integrations, including Google Ads API (ConversionUploadService, Enhanced Conversions), Meta Conversions API (CAPI), or similar server-side event tracking systems.
•Knowledge of identity resolution and entity matching techniques, including graph-based identity stitching, probabilistic matching, or cross-device identity resolution.
•Experience with Terraform, CI/CD pipeline design, and Docker containerization for ML workloads.
•Knowledge of causal inference methods, A/B testing in advertising environments, or multi-armed bandit optimization.
•Familiarity with AI-assisted development tools and agentic coding workflows, such as Claude Code, Cursor, or similar solutions.
Условия
•Opportunity to attend and participate in the company's BA meetups, as well as a compensation program for international professional certificates.
•Many different ways to grow and develop at the company: you can improve as a tech specialist or a resource manager, and all your activities will be decently rewarded.
•Opportunity to take on the role of a mentor and help a group of BA/SA beginners grow.
•Access to Andersen Knowledge Base, with courses in a variety of areas: the art of negotiation, PreSale & Discovery, Agile essentials, product ownership, microservice architecture, API design, SQL, AWS, GDPR, UX/UI, people management, and more.
