Maxim Werachowski
Senior Java Software Engineer @Neo Group
Signup · Get unlimited contacts
WORK HISTORY
Senior Java Software Engineer @Neo Group
Tech lead at a real-money sports-betting and gaming operator on GKE, 100+ microservices. Lead a team of five backend engineers- Scaled event-processing throughput 6x and cut p99 latency ~30% via Kafka tuning (compression, partitioning, consumer balancing), distributed Hazelcast caching, PostgreSQL optimization, and CQRS read/write split on hot paths- Designed the new platform\'s event-driven messaging on Kafka, not extending the legacy RabbitMQ stack-avoided its in-memory and parallel-processing constraints at 100+ microservice scale- Built a production AI/LLM pipeline for sports-entity deduplication and multilingual translation across ~1M competitors and markets (~100-200 new/day). Tiered matching:~60% deterministic (pg_trgm+normalization),~15% AI-resolved (pgvector embeddings + LLM reasoning),~25% to traders.~2% false-positive rate on AI suggestions, tuned via prompt engineering and name normalization. GPT-4.1 for translations. Ollama in dev, OpenAI (text-embedding-3-small+GPT-4.1) in prod. Duplicates reaching prod dropped to 2-3/week- Instrumented the AI pipeline end-to-end: custom service metrics→Prometheus→Grafana dashboards tracking every trader and AI action- Architected the real-time WebSocket layer powering live odds, bets, and player actions. Replaced NewRelic with a self-hosted Grafana+OpenSearch+Prometheus stack on GKE, cutting tooling spend- Personal AI-augmented engineering workflow: agentic coding via Claude and OpenAI Codex, task patterns for agents. Rolled out AI-assisted code review across the team- Tech Lead and quality: hiring, onboarding, architecture reviews, performance and career conversations, mentoring. Raised test coverage to 70%, rolled out TDD, established code review standards and coding conventions.Stack: Java 21, Java 25, Spring Boot 3, Kafka, PostgreSQL, Hazelcast, Redis, GKE, Kubernetes, Docker, WebSockets, Spring AI, OpenAI API, Ollama, pgvector, pg_trgm, OpenSearch, Grafana, CQRS, JUnit/Mockito.
EDUCATION
Vitebsk State University named after P.M. Masherov
Bachelor, Law, general direction
Vitebsk State University named after P.M. Masherov
Bachelor, Mathematics and Computer Science
ABOUT MAXIM WERACHOWSKI
Senior backend engineer with 13+ years building high-load distributed systems. Currently at a real-money sports-betting and gaming operator running 100+ microservices on GKE - own architecture across the platform and lead a small backend team.What I bring to the table: High-load scaling you can measure. At my current company I scaled event-processing throughput 6x while cutting p99 latency by ~30%- through Kafka tuning, distributed Hazelcast caching, CQRS, and PostgreSQL optimization. Not by throwing hardware at the problem.Production AI/LLM, not just demos. Built a deduplication and translation pipeline for ~1M sports entities combining pgvector embeddings, PostgreSQL trigram matching, and GPT-4.1 reasoning. Tiered architecture (deterministic → AI → human) keeps inference cost under control: Ollama in dev/staging, OpenAI in prod. The AI layer ships at ~2% false-positive rate, continuously measured and tuned under live traffic.Cost-aware infrastructure decisions. Replaced NewRelic with a self-hosted Grafana + OpenSearch + Prometheus stack on GKE; chose Ollama over OpenAI in non-prod. I treat cloud and tooling spend as part of the architecture, not someone else\'s problem.Engineering experience that goes beyond the IC seat. Lead a backend team of five — hiring, mentoring, architecture and design reviews — while still shipping code daily. Senior IC is what I\'m looking for next.Core stack: Java 21 / Java 25, Spring Boot 3, Kafka, PostgreSQL, Kubernetes (GKE), Docker, Hazelcast, Redis, WebSockets, Spring AI, OpenAI, Ollama, pgvector, pg_trgm, Grafana, Prometheus, OpenSearch, CQRS, Event-Driven, I add the most value: products that outgrew their first backend — latency spikes, deploy fear, on-call fatigue, a queueing layer bottlenecking the rest of the platform. Or product teams shipping their first real LLM feature and discovering that the prototype doesn\'t survive under real load. I\'ve turned both situations around.Logistics: based in Berlin · No visa, sponsorship, or work permit needed. English B2 (professional working proficiency). German B1+(daily use in Berlin). Open to on-site, hybrid, or remote in Berlin / Germany.Open to: Senior Backend Engineer roles in Berlin / Germany. DMs are open.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.