At Yandex, I designed, implemented, and launched a service-document processing backend using S3, PostgreSQL, and Kafka. Its digitized records feed vehicle-maintenance recommendations.
I added ML processing that handled 95% of an average 30,000 documents per month over nine months. For ML-processed documents, delivery time fell from about an hour to a minute, with estimated recognition-cost savings of US$920K against manual processing of the same volume.
I migrated a domain from a shared REST proxy to GraphQL with one other engineer, preserving backward compatibility and enabling independent deployments. Development cycle time fell 54% for proxy-dependent tasks and 18% across team tasks.
I also worked with the team on monitoring, testing, and technical debt; service-unavailability incidents fell from about five per quarter to zero over six consecutive quarters. At T-Bank, I built and launched the first integration with T-Remont.
