How Nozir Shokirov Uses Operations Research and Data Science to Optimize Shipment Consolidation

His focus? Developing efficient, data-driven algorithms that optimize processes and empower decision-makers to make the best possible decisions.
 

Logistics Matters: Can you describe the role of an Operations Research Specialist?

Nozir Shokirov: An Operations Research Specialist (ORS) applies advanced mathematical techniques to optimize business processes by developing solutions that facilitate objective and data-driven decision-making. These decisions can range from strategic choices, such as determining the location of a new terminal, to operational decisions, like the daily process of scheduling and routing orders. As ORSs, our role is to provide solutions that yield the best possible decisions while considering relevant business and technical constraints. We achieve this by translating business needs into mathematical models and developing cutting-edge algorithms that scale efficiently.

Typically, we need to look several layers beyond the initial business requirements when developing solutions, because some problems that appear distinct on the surface can be mathematically mapped to the same underlying problem. This procedure allows us to develop broader solutions that can address multiple scenarios without the need to completely reinvent the wheel. More importantly, it empowers us to consistently maintain the high standards of our solutions and significantly speeds up our development time.

For instance, consider two seemingly unrelated scenarios. First, imagine you are going on a vacation, and you have several items to pack into your suitcases. Each suitcase has a weight limit (capacity constraint). Your goal is to fit all the items into the fewest suitcases possible. In the second scenario, imagine you are redecorating your room, and you have rolls of wallpaper. Each roll has a fixed length, and you need to cover the entire room without wasting any wallpaper. Although these problems appear different at first, mathematically, they share common ground as both involve similar constraints (e.g., suitcase weight limits or fixed-length wallpaper rolls) and optimization goals (fitting items efficiently). As ORSs, we recognize this similarity and create a unified solution applicable to both scenarios. Also, if you think about it, these hypothetical scenarios are related to practical applications that are part of DB SCHENKER business, such as optimally arranging cargo in shipping containers or efficiently using storage spaces at our sites. They fall under the broader category of packing problems, which our OR team effectively tackles with the state-of-the-art in-house solutions we develop.
 

Logistics Matters: Operations Research is a broad field – Can you describe a typical team structure and how different teams in Operations Research work together?

Nozir Shokirov: Our Operations Research (OR) team is organized into Competence Centers (CCs) based on our expertise and the business problems we address. Within each CC, we have OR experts with deep technical knowledge of OR techniques relevant to specific business challenges. This structure paves the way for us to build invaluable and specialized domain expertise within the relevant processes.

Additionally, the ORSs collaborate closely with Business Consultants (BCs), who serve as product owners and assist in translating the business requirements. Notably, our BCs also possess in-depth knowledge of OR techniques, significantly enhancing our collaboration.

While each CC primarily focuses on projects within its specific domain, some commonalities cut across CC boundaries. We develop solutions that are business-agnostic, necessitating synergy between the CCs. Frequent sharing of best practices also ensures consistency and effectiveness.

To stay at the forefront of modern OR practices, we actively participate in leading scientific and logistics-specific conferences. This allows us to continually push the boundaries of state-of-the-art methods while addressing the complex logistic problems we tackle.
 

Logistics Matters: How do you leverage data and algorithms in your role as an Operations Research Specialist at DB SCHENKER?

Nozir Shokirov: As ORSs, we develop algorithms and solutions to real-world business problems, so using the right data and making data-driven decisions is at the core of these solutions. In this process, we collaborate with professionals from various departments, including Data Engineers who construct data pipelines for data acquisition, Platform Engineers who assess the technical feasibility of diverse solutions, and Backend and Frontend Developers who integrate our solutions into the broader IT ecosystem of DB SCHENKER.

Our projects involve developing algorithms that support the decision-making process rather than replacing it. Therefore, a key stakeholder in our projects is the individuals who make these decisions. Their invaluable feedback and domain expertise significantly influence the development and direction of our algorithms. For instance, a project I am involved in, alongside my colleagues in the Routing CC, focuses on how to effectively consolidate shipments from different branches so that the overall transportation cost of delivering these shipments is minimized. The goal is to provide tour suggestions to dispatchers, assisting them in their routing and scheduling operations. From the onset, we involve the dispatchers in the process and hold regular feedback sessions to discuss our algorithm’s results and continually enhance it.

In terms of algorithm development, as mentioned before, we diligently look for ways to identify common elements in our projects and develop generic structures that form the building blocks of our algorithms. For example, every project has a business constraint that physically restricts the resources that can be utilized. This could be the size of a truck for loading shipments, the utilization of storage space in a terminal, or the volume of a cargo plane or shipping container used for transporting items. Instead of case-specific static implementation, we develop a generic dynamic data structure called “quantities” which can handle several dimensions and be used in all cases.

Another crucial aspect is that we thoroughly test everything we develop, including these building blocks, which are also continuously improved. This strategy enables our relatively small team to efficiently develop robust and reliable solutions at a rapid pace.
 

Logistics Matters: How do you envision the collaboration between Operations Research and AI algorithms in optimizing logistics processes?

Nozir Shokirov: I believe that AI and OR are two complementary fields with immense potential for optimizing logistics processes when combined. OR harbors powerful techniques with a strong theoretical foundation behind them and when effectively integrated with the adaptability and predictive capabilities of AI algorithms, they can efficiently address complex logistics challenges.

For example, let us get back to the issue of leveraging data to make more informed decisions. We can construct pipelines where the necessary data is filtered, cleaned, or even generated using AI models. This data is then fed into an optimization framework where robust OR models solve an optimization problem. The solution can then be utilized in the decision-making process. The entire process can be monitored by another AI algorithm, which prepares new data and triggers the OR models when changes occur in the environment. This approach enhances the system’s robustness to the ever-changing processes in logistics.

Another potential use case involves leveraging large language models (LLMs) to translate business requirements into mathematical models. These LLMs excel in processing documents, extracting key information from domain-specific text, and can act as intermediaries in translating business requirements into mathematical models for optimizing logistic processes.

There are numerous other potentials where AI predictions can inform OR models, improving their effectiveness and vice versa OR models can guide AI algorithms to generate more interpretable and actionable results. This constructive collaboration, with OR providing the mathematical backbone and AI contributing adaptability, data-driven insight, and automation, offers the possibility of producing remarkable results in optimizing logistics operations.

This collaboration allows the field of OR to benefit from the popularity of AI, increasing awareness of these powerful techniques for solving complex problems.