From Algorithms
to Impact

At Amazon, I manage products focused on optimizing the Amazon Fulfillment Network. This involves restructuring the US network based on future demand forecasts, partner contracts, and build plans for new facilities (Fulfillment Centers, Sortation Centers, Delivery Stations). Our goals are to delight customers with 1-2 day delivery while minimizing operational costs.

I oversee the development of scientific models and model cascades for prediction (Machine Learning) and optimization, aligned with business objectives. I also manage the scientific platform where these models operate, enabling business users to run simulations and scenario analyses. Collaborating with internal stakeholders, I identify problems to translate them into product requirements, working closely with engineers, analysts, and scientists.

My products played a crucial role in enabling the regionalization of the delivery network. This initiative reduced the overall complexity of the network by creating distinct regions, which simplified operations and allowed for more efficient inventory placement strategies. As a result, we were able to prioritize in-region deliveries, leading to faster delivery times and optimized operational efficiency.

At Auctane, I was responsible for the evolution of our integration platform, ensuring better internal adoption across brands like ShipStation, Metapack, Stamps.com, and Shipworks. My role involved enhancing the platform’s capabilities to support a wider range of carrier and order source features, such as those from Shopify and Magento, to improve overall functionality and user experience.

In addition to this, I managed the Amazon Marketplace integration module. This involved incorporating advanced features from Amazon's API into our platform and guiding customers through the adoption process. I ensured that the Amazon integration module was fully utilized by our brands, optimizing the shipping experience for our customers. My work impacted over 20,000 merchants using these integrations, contributing to an annual shipment volume of 150M+ across all integrations I managed.

I led a team of 40 Individual Contributors, including developers, analysts, product managers, and QA engineers, to create a comprehensive food delivery service for Azbuka Vkusa. My responsibilities included assembling the team and approving the initial architecture, with a focus on fault tolerance, scalability, and privacy protection. I made key decisions on the technological solutions for various system components, ensuring the system was built on a robust, scalable foundation.

We developed a system that managed everything from order processing and SKU management to payments and last-mile delivery. Our scalable system integrated seamlessly with web and mobile applications, ensuring customers could easily access high-quality, ready-made meals delivered within 30-60 minutes. The system was also designed to be resilient to peak demand spikes, maintaining consistent performance during high-traffic periods.

To enhance the user experience, we implemented intelligent routing algorithms that optimized delivery speed and cost by selecting the most efficient routes and courier services. We also developed integrations with partner services, including third-party food delivery companies, to balance the load on our internal delivery service. Mobile tools were created for couriers and order pickers, enabling end-to-end process control and real-time data collection for targeted optimizations. The service was successfully launched and scaled to cover new geographies, significantly enhancing convenience and efficiency in food delivery.

I led the development of a personalization service aimed at enhancing the omnichannel customer experience for Azbuka Vkusa, particularly in the ready-made food segment. The project’s goal was to personalize the shopping experience across both online platforms—websites and mobile apps—and offline channels.

For guest users, I developed models that could instantly update a user’s profile and dynamically rank the product catalog, including up-sell and cross-sell suggestions, within milliseconds. For returning customers, especially those with accounts and offline purchase histories, I created machine learning models that tailored product recommendations and special offers based on their shopping behavior.

To extend personalization into the offline experience, we developed APIs that allowed personalized push notifications about discounts and offers available at nearby stores. This ensured a consistent and tailored experience across all customer touchpoints.
I served multiple roles in this project, acting as a Machine Learning Engineer, Product Manager, and Engineering Manager. I developed the initial models, assembled and managed a team of 10, and oversaw the transition of these models into production. One of the key technical achievements was reducing response times to under 0.15 milliseconds for the 95th percentile of users, ensuring quick and smooth interactions.

We launched the service for a limited user base to gather data and refine our approach, focusing on segment-specific models, such as those for vegetarians and families, to further personalize the experience. My main objectives were to reduce the time from site entry to purchase and to increase the average order value through effective personalization.