Roman Seleznev

With over 20 years of experience in the IT industry, I’ve come to understand that technologies should provide tangible value by solving real-world problems and bringing happiness to its users. This understanding guides me in creating meaningful solutions across various domains. In this nomination, I present a technological innovation that enables differentiated access management based on the importance of consumers. As a result, it improves the overall quality and stability of IT solutions.

In 2022, I designed and launched a solution at Sberbank (the largest universal bank in Russia and Eastern Europe) that enabled clients to automatically receive tax refunds, reducing the waiting period from four months to just one month. This solution was awarded the Digital Leaders Award 2022 in the category “Solution of the Year” within the investment banking services sector.

In 2023, I designed and introduced new features in Sberbank’s mobile app that allow users to perform financial transactions using voice commands. This innovation won the Investment Leaders Award in the “Online Assistant of the Year” category for informational support in investing, as well as the National Banking Award in the “Investment Product” category.

In 2024, during my personal research, I delved deeper into system design principles. Through this exploration, I identified gaps in existing rate-limiting methodologies. Consequently, I formulated and experimentally validated new methods addressing these deficiencies. Additionally, in 2024, I presented at the prominent Code Fest conference, where I shared my original approaches to improving the analysis and design of information systems.

This comprehensive track record underscores my commitment to driving technological innovation and creating impactful solutions across industries.

At the core of my focus lies the client and their needs. This approach ensures that I avoid creating overly complex yet unnecessary solutions. Additionally, the innovations I develop often reveal untapped potential for further improvements, allowing me to remain optimistic about future possibilities.

Companies today rely heavily on distributed systems and intricate ecosystems composed of numerous interconnected information systems. These systems interact through APIs, delivering vital data and functionalities required for seamless business operations. However, unchecked access to these resources may lead to escalating operational costs, overburdened servers, degraded performance, and even catastrophic system failures. These challenges frequently stem from abuse by specific users or malevolent actions such as DDoS attacks.

In response to these threats, rate-limiting algorithms have been developed. At the same time, existing algorithms do not differentiate between types of users, leading to scenarios where lower-priority users exhaust available quotas, depriving critical users of essential services. For example, free-tier or unregistered users could monopolize all available resources, leaving paying subscribers without access. In e-commerce, this issue manifests as heightened infrastructure costs and potential attrition of premium clients. Similarly, in digital marketing, it generates inaccuracies in analytics due to duplicate requests and disrupted communication between advertisers and platforms.

I devised two novel methods based on the time-tested token-bucket algorithm to address these shortcomings. By implementing a classification framework for API consumers, my methods guarantee equitable allocation of resources among disparate user groups while ensuring consistent access for high-value customers. These methods not only boost system efficiency but also bolster customer satisfaction by curtailing service outages and enhancing overall dependability. Furthermore, they empower businesses to exercise greater control over operational expenditures by mitigating superfluous resource usage.

It’s worth noting that public literature on rate limiting has been sparse until now. My methods were recently published in a scientific journal online, making them accessible to anyone who wishes to adopt and build solutions based on them, thus increasing their operational effectiveness.