Building robust and scalable machine learning and optimization systems that drive business impact.
I am a Machine Learning, Optimization, and MLOps engineer specializing in designing, deploying, and optimizing intelligent decision-making systems. With a passion for principled approaches to transforming data into decisions, I help organizations unlock value through automation, predictive analytics, optimization, and robust architectures.
Skills: Problem framing, experimental design, productionization, federated learning, supervised and unsupervised learning
Tech: Python, Flower, AutoML, scikit-learn, SparkML, Pytorch, XGBoost, LightGBM, SHAP, Optuna
Skills: CI/CD/CT, IaC, drift detection and handling, code, data, and artifacts version control, automated testing, alerting, monitoring, experiment tracking
Tech: GitHub Actions, Terraform, Lakehouse Monitoring, Unity Catalog, MLFlow, Weights & Biases, pytest, SonarQube
Skills: ETL pipelines, Data Governance
Tech: Data Factory, Databricks jobs, Spark, Unity Catalog
Problem: Customers are unsatisfied with startup times of their infotainment systems.
Solution: Developed a predictive system identifying user routines to quickstart the infotainment system, integrated into enterprise architecture.
Outcome: Improved user experience for >>100k customers.
Tech Stack: Azure, Databricks, Data Factory, Azure API Management, Spark, Python, Terraform, SQL
View official VW communicationProblem: Manual deployment, testing, and operations are time-consuming and error-prone.
Solution: Built a CI/CD/CT-enabled MLOps platform with automated testing, monitoring, and alerting.
Outcome: Reduced model update time from (sometimes) days to minutes; reduced time until issues are noticed.
Tech Stack: Azure, Databricks, lakehouse monitoring, MLflow, model serving endpoints, GitHub Actions, Terraform, Docker
Problem: Product usage data is underutilized in product strategy decisions.
Solution: Built and integrated an unsupervised machine learning system identifying distinct groups of product usage behavior.
Outcome: Increased data utilization with a positive impact on product decisions.
Tech Stack: Azure, Databricks, MLflow, Spark, SparkML
Problem: Manual evaluation of fiber expansion areas can be time-consuming.
Solution: Designed and implemented a decision-support tool based on linear programming and heuristics.
Outcome: Reduced evaluation times from days to hours.
Tech Stack: GCP, Python, Gurobi, Docker