Senior Software Engineer / Team Lead
Germany
Alexey
Zabegin
Council Member
Alexey Zabegin is a Senior Software Engineer and Team Lead with more than 9 years of experience designing, building, and operating scalable, high-performance distributed systems. He specializes in Java, Spring Boot, PostgreSQL, Kafka, and microservices architecture, and his work spans fintech and consumer platforms. He has delivered production backend systems serving 30M+ users, optimized PostgreSQL workloads across 500M+ records, and built financial integration platforms processing 20M+ messages per day.
Since November 2025, he has served as Senior Software Engineer at Haut AI, a B2B SaaS company focused on AI skincare analysis. He architects the multi-tenant backend behind the real-time Skin.Chat AI skincare advisor, which serves 60 beauty brands. He designed the core service layer that enables per-brand model configuration and personalized recommendations at production scale. Because no reusable AI reasoning layer existed across brand integrations, he built end-to-end agentic pipelines that deliver context-aware recommendations at sub-200ms latency, which accelerated feature delivery 2x for a platform serving 10K+ daily active users. He also implemented a full LLM observability stack with Langfuse across 3+ production environments, tracking prompt performance, token usage, and model quality per brand. This reduced debugging time by 30% and enabled continuous eval-driven prompt iteration. In addition, he builds multimodal pipelines that extract structured data from mixed image and text documents, covering product label parsing and ingredient analysis across 60+ brand catalogs.
From October 2023 to October 2025, he was Senior Software Engineer at Replika AI, a B2C AI companion platform with more than 30M users. He redesigned and optimized high-traffic backend APIs, introducing caching and request optimization strategies that brought response latency under 2 seconds across endpoints. He architected a persistent memory service in Java and PostgreSQL, using table partitioning and query optimization to manage 500M+ records while keeping p99 query latency below 15ms. He migrated synchronous notification processing to an asynchronous Kafka-based architecture with Java consumers, which decoupled services, eliminated blocking operations, and raised the notification delivery rate to 99%. He also refactored core user cohort processing algorithms for batches of 500 to 1,000 users, removing redundant computations and achieving a 2x speed improvement. He designed and implemented a feature flag system that supports granular feature exposure and progressive rollouts, reducing deployment blast radius and making production releases safer.
Since November 2025, he has served as Senior Software Engineer at Haut AI, a B2B SaaS company focused on AI skincare analysis. He architects the multi-tenant backend behind the real-time Skin.Chat AI skincare advisor, which serves 60 beauty brands. He designed the core service layer that enables per-brand model configuration and personalized recommendations at production scale. Because no reusable AI reasoning layer existed across brand integrations, he built end-to-end agentic pipelines that deliver context-aware recommendations at sub-200ms latency, which accelerated feature delivery 2x for a platform serving 10K+ daily active users. He also implemented a full LLM observability stack with Langfuse across 3+ production environments, tracking prompt performance, token usage, and model quality per brand. This reduced debugging time by 30% and enabled continuous eval-driven prompt iteration. In addition, he builds multimodal pipelines that extract structured data from mixed image and text documents, covering product label parsing and ingredient analysis across 60+ brand catalogs.
From October 2023 to October 2025, he was Senior Software Engineer at Replika AI, a B2C AI companion platform with more than 30M users. He redesigned and optimized high-traffic backend APIs, introducing caching and request optimization strategies that brought response latency under 2 seconds across endpoints. He architected a persistent memory service in Java and PostgreSQL, using table partitioning and query optimization to manage 500M+ records while keeping p99 query latency below 15ms. He migrated synchronous notification processing to an asynchronous Kafka-based architecture with Java consumers, which decoupled services, eliminated blocking operations, and raised the notification delivery rate to 99%. He also refactored core user cohort processing algorithms for batches of 500 to 1,000 users, removing redundant computations and achieving a 2x speed improvement. He designed and implemented a feature flag system that supports granular feature exposure and progressive rollouts, reducing deployment blast radius and making production releases safer.
Between March 2018 and October 2023, he worked as Team Lead at Sberbank, a top-3 bank by assets, on the HR SaaS platform Pulse, which handles 50K+ resumes per month. HR analysts had been cross-referencing candidate data across separate systems by hand, so he designed a unified data view that halved candidate analysis time and doubled throughput. His work on recruitment operations cut time-to-hire by 90% and doubled overall candidate quality scores. He integrated an LLM model to process and analyze 50K+ resumes monthly, extracting key skills and experience and matching candidates to open positions with 85%+ accuracy. He also designed and implemented an intelligent candidate screening service for enterprises with 500+ employees, reducing manual screening time by up to 90%. He developed relevance scoring algorithms that ranked and shortlisted candidates automatically, increasing HR shortlist precision by 95%. He implemented secure data handling and compliance measures for sensitive personal information, ensuring 100% compliance.
From September 2014 to March 2018, he was Senior Software Engineer and Team Lead at Sber-Tech, where he worked on the Global Market Service Bus. He designed and architected a Spring Boot and Spring Cloud integration platform, first deployed as Linux services on virtual machines, and delivered it from concept to production as a reusable foundation for enterprise service integrations. Business developers had been blocked by low-level infrastructure complexity on every new integration, so he built core platform components that abstracted those details, shortened time-to-integration, and separated domain logic from infrastructure concerns. He led the migration of the platform from virtual machines to OpenShift, redesigning the architecture for fault tolerance, horizontal scalability, and operational resilience, and the platform supported production workloads of 20M+ messages per day. He also created reusable microservice templates that standardized common infrastructure components, reduced code duplication, and sped up new service development by 1.5x. Other deliverables included an end-to-end integration that transfers financial data from Bloomberg to internal downstream systems, and a database migration from Oracle to PostgreSQL.
Alexey Zabegin holds a Master's degree in applied management and informatics in technical systems from the University of Instrument Engineering and Computer Sciences, completed in 2009. He has also earned certifications in Jenkins, Sonar, and Ansible through Sber-Tech (2016), Spring Framework through Udemy (2016), and Apache Camel through Udemy (2018).
His technical expertise covers Java, Kotlin, Python, Spring Boot, Spring AI, Spring Cloud, REST and GraphQL APIs, Apache Camel, Kafka, MQ, Apache Qpid, jOOQ, MyBatis, Spring JDBC, JUnit, Mockito, LangChain, LangGraph, and Pydantic. He works with PostgreSQL, Oracle, MongoDB, and Redis, and with Keycloak, Linux, ELK, OpenSearch, Prometheus, Grafana, and Datadog. His infrastructure experience includes Docker, Kubernetes, Helm, Ansible, AWS, GCP, Azure, OpenShift, GitLab, Bitbucket, Jenkins, and CI/CD pipelines. In AI engineering, he works with agentic AI, RAG, LLM evaluation and observability, evals infrastructure, prompt engineering, embeddings, and structured outputs.
From September 2014 to March 2018, he was Senior Software Engineer and Team Lead at Sber-Tech, where he worked on the Global Market Service Bus. He designed and architected a Spring Boot and Spring Cloud integration platform, first deployed as Linux services on virtual machines, and delivered it from concept to production as a reusable foundation for enterprise service integrations. Business developers had been blocked by low-level infrastructure complexity on every new integration, so he built core platform components that abstracted those details, shortened time-to-integration, and separated domain logic from infrastructure concerns. He led the migration of the platform from virtual machines to OpenShift, redesigning the architecture for fault tolerance, horizontal scalability, and operational resilience, and the platform supported production workloads of 20M+ messages per day. He also created reusable microservice templates that standardized common infrastructure components, reduced code duplication, and sped up new service development by 1.5x. Other deliverables included an end-to-end integration that transfers financial data from Bloomberg to internal downstream systems, and a database migration from Oracle to PostgreSQL.
Alexey Zabegin holds a Master's degree in applied management and informatics in technical systems from the University of Instrument Engineering and Computer Sciences, completed in 2009. He has also earned certifications in Jenkins, Sonar, and Ansible through Sber-Tech (2016), Spring Framework through Udemy (2016), and Apache Camel through Udemy (2018).
His technical expertise covers Java, Kotlin, Python, Spring Boot, Spring AI, Spring Cloud, REST and GraphQL APIs, Apache Camel, Kafka, MQ, Apache Qpid, jOOQ, MyBatis, Spring JDBC, JUnit, Mockito, LangChain, LangGraph, and Pydantic. He works with PostgreSQL, Oracle, MongoDB, and Redis, and with Keycloak, Linux, ELK, OpenSearch, Prometheus, Grafana, and Datadog. His infrastructure experience includes Docker, Kubernetes, Helm, Ansible, AWS, GCP, Azure, OpenShift, GitLab, Bitbucket, Jenkins, and CI/CD pipelines. In AI engineering, he works with agentic AI, RAG, LLM evaluation and observability, evals infrastructure, prompt engineering, embeddings, and structured outputs.