Volodymyr Koliadin

Volodymyr is a Senior Data Scientist at Grid Dynamics. He joined Grid Dynamics in 2021 and has more than 15 years of experience working on data driven solutions.

Volodymyr Koliadin

Optimization of order and inventory sourcing decisions in supply chains with multiple nodes, carriers, shipment options, and products

This article discusses the challenges of optimizing order sourcing decisions in complex logistics environments and compares regular optimization methods (such as mixed integer programming) with metaheuristic (stochastic) optimization methods. It provides insights into the advantages and limitations of each approach and highlights the scalability and flexibility of metaheuristic algorithms for high-dimensional problems.

Building a predictive maintenance solution using AWS AutoML and no-code tools

AWS has outlined how equipment operators can build a predictive maintenance solution using AutoML and no-code tools provided by the company. The solution uses machine learning techniques to estimate the remaining useful life (RUL) of machines or equipment, allowing operators to optimise maintenance schedules and balance resource usage and failure risks. The solution can be implemented using AWS Canvas or AWS AutoPilot, depending on the specific requirements of the application.

Anomaly detection in industrial IoT data using Google Vertex AI: A reference notebook

This blog post discusses the challenges of IoT data analysis for system health monitoring and provides a reference pipeline for anomaly detection using machine learning techniques. The pipeline includes training regression models, computing anomaly scores, and making binary decisions to detect anomalies in IoT data.

Detecting anomalies in high-dimensional IoT data using hierarchical decomposition and one-class learning

This article discusses a methodology for designing machine learning-based health monitoring systems for complex industrial systems. It emphasizes the use of hierarchical decomposition and one-class learning to address challenges such as high dimensionality, high data rates, and qualitative and quantitative inhomogeneity of sensor readings.

Anomaly detection in industrial applications: Solution design methodology

This article discusses the importance of anomaly detection in technical systems and outlines a solution design methodology based on the types and availability of labeled data. It highlights the pitfalls of using unsupervised methods and recommends the use of one-class learning approaches, even in situations where two-class labeling is available or no labeled data is present.