Yevhenii Orlov

• Designed and implemented a BYOM (Bring Your Own Model) platform to enable dynamic onboarding of PMML models into the company’s infrastructure. The platform maps model features to the company’s Snowflake Big Data environment, deploys models using SageMaker, and establishes automated feature engineering processes to ensure up-to-date feature availability for inference. Developed a batch scoring system integrated with the frontend to present real-time predictions alongside Big Data insights.

• Optimized ML model training processes, reducing the training time of the company’s most successful regularly-trained model by 80% and associated costs by 75%, significantly improving operational efficiency.

• Acted as a Kubernetes expert, spearheading strategic migrations from EC2/ECS to EKS across departments. Delivered lectures on Kubernetes technologies, including Helm, Kustomize, and cloud-native workflows, fostering organization-wide adoption of best practices in container orchestration.

• Led the development of the MLOps Workbench project, building a robust, integrated platform to empower data scientists with seamless MLOps capabilities and Big Data integration, enhancing productivity and operational excellence.

• Developed an enterprise-grade Kubeflow distribution, collaborating with the open-source community to customize and adapt Kubeflow for OpenShift as an installation option, aligning with enterprise deployment needs.

• Built a high-demand Remote Lab (Notebooks) product, featuring user and role-based permission management on Kubernetes with seamless SSO integration. Integrated the product with other privately deployed frameworks, including Spark, MLflow, Feast, and Airflow, creating a unified and secure data science platform.

• Customized the KServe component of Kubeflow to support a multi-user scenario, enabling seamless model serving integration with the Remote Lab (Notebooks) product, ensuring secure and efficient operations in shared Kubernetes environments.

• Designed and implemented ETL and ML training pipelines using Apache Spark, tailored to meet diverse customer requirements, optimizing performance and scalability.

• Developed cloud-native solutions by creating Golang operators, enhancing the automation and manageability of platform components in Kubernetes environments.