ML/DS Engineer (Classic ML)

Remote ML/DS Engineer position at Noveradev, developing classic ML models (gradient boosting, logistic regression) for credit scoring, churn prediction, risk assessment, and recommendation systems.

Develop and implement classic machine learning models: gradient boosting (LightGBM, XGBoost, CatBoost), logistic regression, random forest, linear models. Solve business tasks: credit scoring, churn prediction, risk assessment, recommendation systems based on collaborative filtering, time series forecasting. Feature engineering – creation and selection of features, work with large volumes of structured data (SQL, Spark, Pandas). Build training and validation pipelines considering interpretability requirements (SHAP, LIME, Feature Importance). Develop pipelines for A/B testing and post-release model monitoring (data drift, target variable drift). Integrate models into production environment via API (FastAPI, Flask) and orchestration (Docker, Kubernetes). Interact with clients: clarify business metrics, present results, write technical documentation.

Commercial experience in Data Science / ML. Deep knowledge of Python and stack: pandas, numpy, scikit-learn, and one of the gradient boosting frameworks (LightGBM, XGBoost, CatBoost). Confident SQL – writing complex queries, working with big data (ClickHouse, PostgreSQL, Greenplum or similar). Understanding quality metrics for classification, regression, ranking: precision/recall, ROC-AUC, lift, MAP, NDCG. Experience building end-to-end ML pipelines: from data extraction to deployment and monitoring. Knowledge of MLOps tools: MLflow, DVC, experience with CI/CD for ML projects. Nice to have: experience with uplift modeling (causal models, Meta-learners, decision tree-based uplift models); experience in credit scoring – building scorecards, working with credit bureaus, regulatory requirements for interpretability; practical experience with AI tools (Claude Code, OpenAI Codex, OpenCode) in commercial development – for code, tests, reviews, documentation – with mandatory result verification and confidentiality compliance.