Senior Data Scientist

Senior Data Scientist

At Global IT Hub, we are more than a technology organization. We are a community of innovators, problem-solvers, and digital pioneers dedicated to accelerating Atlas Copco Group’s business potential through world-class technology solutions. Join us to shape the future of digital transformation, work on global challenges, and create impact for businesses and customers worldwide.

CLASSICAL MACHINE LEARNING | JOB DESCRIPTION

Role purpose Lead the design, validation, deployment and lifecycle ownership of production-grade classical machine learning solutions that convert business problems into measurable outcomes. The role combines deep algorithmic expertise, disciplined CRISP-DM execution, MLOps practices and senior technical leadership.

Role at a glance

Team

AI Team, Global IT Hub

Primary stakeholders

Business product owners, IT, data engineering, analytics and domain teams

Experience

5+ years in data science / machine learning, including 3+ years owning production ML solutions

Education

Bachelor’s or Master’s degree in Data Science, Statistics, Mathematics, Computer Science, Engineering or a related quantitative field

Key responsibilities

• Partner with stakeholders to frame ambiguous business needs as well-defined machine learning problems, including decision context, baselines, constraints, risks and measurable success criteria.
• Lead end-to-end delivery using CRISP-DM: business understanding, data understanding, data preparation, modeling, evaluation and deployment, with iteration and traceability across phases.
• Design, train and compare classical supervised and unsupervised learning models; select the simplest approach that meets performance, interpretability, latency, scalability and cost requirements.
• Own feature engineering, data leakage prevention, sampling strategy, cross-validation, hyperparameter optimization, threshold selection, calibration and robust error analysis.
• Build reliable evaluation frameworks using fit-for-purpose metrics, statistical tests, confidence intervals, business-cost functions and champion-challenger comparisons.
• Take models from experimentation to production with engineering teams, covering packaging, APIs or batch scoring, versioning, CI/CD, monitoring, retraining and rollback.
• Own post-production performance, including drift detection, data-quality controls, incident analysis, model refresh decisions and communication of model limitations.
• Apply explainability and responsible AI practices through model cards, reproducibility, bias and fairness checks, audit trails, privacy-aware design and human oversight.
• Set technical standards, conduct design and code reviews, mentor data scientists, create reusable components and influence the classical ML roadmap.
• Communicate recommendations, uncertainty, assumptions and trade-offs clearly to technical teams, business leaders and governance forums.

Mandatory Technical Skills

• Advanced Python and SQL, with strong hands-on use of NumPy, Pandas, scikit-learn, SciPy and statsmodels; ability to write modular, tested and maintainable code.
• Deep knowledge of regression and classification: linear and logistic regression, regularization, decision trees, Random Forest, Extra Trees, XGBoost, LightGBM, CatBoost, support vector machines, k-nearest neighbors, Naive Bayes and appropriate ensemble methods.
• Strong understanding of unsupervised learning: k-means and hierarchical clustering, DBSCAN, PCA and other dimensionality-reduction methods, anomaly or outlier detection and segmentation evaluation.
• Practical knowledge of forecasting and statistical modeling, including ARIMA/SARIMA, exponential smoothing, trend and seasonality, lag features, back-testing and prediction intervals.
• Expertise in feature selection and engineering, missing-value treatment, categorical encoding, class imbalance, resampling, pipeline design and prevention of target leakage.
• Rigorous evaluation skills across ROC-AUC, PR-AUC, precision, recall, F1, log loss, RMSE, MAE, MAPE/WAPE, lift, gain, calibration and business-impact metrics, selected according to the use case.
• Hands-on optimization using grid, random or Bayesian search, with sound cross-validation design for grouped, temporal and imbalanced datasets.
• Practical MLOps experience with experiment tracking, model registry, source control, automated testing, Docker, CI/CD, deployment and model or data monitoring.
• Working knowledge of explainability techniques such as feature importance, permutation importance, partial dependence and SHAP, including their limitations.
• Experience handling large structured datasets and collaborating on scalable data pipelines in a cloud or distributed environment.

Senior-level capabilities

• Independently owns technical direction and delivery for complex ML use cases from discovery through production adoption.
• Challenges weak problem formulations and prevents unnecessary use of ML when rules, analytics or simpler statistical methods are more suitable.
• Makes defensible trade-offs among accuracy, interpretability, maintainability, latency and cost.
• Reviews experimental design and model evidence before approving production release.
• Mentors team members and raises standards for coding, documentation, reproducibility and stakeholder communication.

Good-to-have experience

• Domain exposure in manufacturing, industrial operations, supply chain, service, pricing, finance, sales or customer analytics.
• Optimization, causal inference, survival analysis, uplift modeling, recommendation systems, geospatial analytics or graph-based techniques.
• Azure Machine Learning, Databricks, Microsoft Fabric, MLflow, Spark and enterprise data sources such as SAP.
• Experience integrating classical ML models into applications, APIs, decision-support workflows or Power BI solutions.
• Awareness of deep learning and generative AI, with the judgment to benchmark them against classical ML rather than defaulting to higher-complexity approaches.

Core algorithm and method coverage

Candidates are expected to demonstrate depth in several categories and sound model-selection judgment across the full landscape. This is not a requirement to have used every library or algorithm in production.

Capability area

Expected knowledge

Supervised learning

Regression, classification, tree ensembles, boosting, SVM, nearest-neighbor and probabilistic baselines.

Unsupervised learning

Clustering, dimensionality reduction, anomaly detection and segmentation validation.

Time-dependent modeling

Forecasting, temporal validation, lag and rolling features, seasonality and uncertainty.

Data and features

EDA, data-quality assessment, transformations, encoding, imbalance handling and leakage control.

Evaluation

Metric selection, baselines, validation design, statistical significance, calibration and business value.

Production lifecycle

Reproducibility, versioning, deployment, observability, drift, retraining and retirement.

Governance

Explainability, documentation, bias and fairness assessment, privacy, approvals and auditability.

What you can expect from us

• Meaningful global business problems and access to cross-functional expertise.
• Freedom to choose fit-for-purpose methods and challenge unnecessary complexity.
• A collaborative environment focused on engineering quality, responsible AI and measurable value.
• Opportunities to mentor, build reusable capabilities and shape the enterprise AI practice.

At Global IT Hub, we believe exceptional talent drives exceptional transformation. As a strategic partner for next-generation digital transformation, we empower our people to innovate boldly, collaborate globally, and turn ideas into solutions that create lasting business value. Join a team where learning never stops, innovation is part of our DNA, and your contribution helps shape the future of a global industry leader.

Role Evolution - Next 3-4 Years

• Evolve from predictive modelling toward decision intelligence, combining prediction, causal analysis, simulation and optimization.
• Build hybrid AI solutions integrating classical ML with Generative AI, foundation models and AI agents.
• Establish systematic benchmarking across classical ML, deep learning and emerging AI models to select the best approach based on performance, explainability, scalability and cost.
• Drive automation of the end-to-end ML lifecycle, including feature engineering, experimentation, deployment, monitoring, drift detection, retraining and model retirement.
• Develop advanced model evaluation frameworks covering technical performance, business impact, robustness, explainability, fairness and operational reliability.
• Advance model observability and continuous learning, proactively managing data drift, concept drift and model degradation.
• Strengthen Responsible AI and model governance, including lineage, reproducibility, explainability, bias assessment, human oversight and auditability.
• Develop reusable ML frameworks, components and engineering standards to accelerate enterprise-wide AI development.