Cosmin Mavrichi

Tech Stack: Java, Spring Boot, BigQuery, GCP Cloud Run, Pub/Sub, Cloud AWS, Monitoring, Cloud Scheduler, PostgreSQL, Kafka, Redis, jOOQ, Micrometer, Auth0, Python, Flyway, GitLab CI • 10M+ events/day at ~8,000 events/sec peak — designed and delivered a GCP-based event diff pipeline that reads BigQuery snapshots, computes field-level diffs, and dispatches batch updates to an external API, including retry strategy and data retention design • ~60% reduction in undetected production incidents — proposed and implemented structured Google Cloud Logging with MDC, GCP log-based alerting, and Micrometer instrumentation across pipeline stages • ~80% reduction in event loss — selected and implemented programmatic Spring Retry with exponential backoff and a Kafka DLQ after evaluating annotation-based alternatives and choosing the more testable design • 4,000+ partner groups served — designed and built a hierarchical configuration system with deep JSON merging, JSON Schema validation, and override flag support, allowing groups to inherit and customize settings independently • Concurrency risks prevented — engineered a scheduled Auth0 organization sync service using ShedLock distributed locking after proactively identifying race-condition risks • Downstream API load reduced — implemented Redis-based caching for Auth0 organization data, improving response times and reducing repeated external API calls • Migration safeguards introduced — led incident response for a high-impact data migration failure, authored the post-mortem, and introduced migration review gates to reduce the risk of repeat data-loss incidents

Tech Stack: Kotlin, Ktor, Java, Spring Boot, Kafka, Elasticsearch, PostgreSQL, Redis, AWS RDS, AWS S3, Azure Blob Storage, Azure SQL Database Database, Terraform, GitLab CI, React • 70% faster re-indexing — identified and resolved a critical Elasticsearch bottleneck, reducing re-indexing time from 5 hours to 1.5 hours by redesigning the indexing strategy • AI-assisted content discovery service built — owned the backend design and integration of a RAG-based service using vector embeddings and Elasticsearch to match relevant video scenes and assets to advertising campaign requirements • Zero-downtime cross-cloud migration — owned the backend migration plan and execution from AWS RDS to Azure SQL Database and from S3 to Azure Blob Storage, coordinating cutover across teams with no data loss • 80% test coverage adopted team-wide — championed stronger unit and integration testing standards across the codebase, reducing regression bugs and CI instability • Secure public APIs shipped — designed and implemented OAuth2/JWT-secured external partner APIs using Spring Security, including API contract design and integration support

Tech Stack: Java, Spring Boot, Hibernate, Spring Data JPA, Oracle SQL, APIGEE, Splunk • New lending APIs delivered — designed and implemented REST endpoints for NatWest's consumer lending platform, taking features from specification through production delivery • ~40% reduction in recurring production defects — investigated critical integration failures, identified root causes, and improved service-level design to prevent repeat incidents• APIGEE policies improved — configured routing, authentication, rate limiting, and request policies for lending platform APIs, improving consistency and reliability across external-facing endpoints Full-Stack Developer Pentalog / Tripadvisor | Mar 2021 – Oct 2021 Tech Stack: Java 11, Spring Boot, PostgreSQL, Kafka, Elasticsearch, Redis, AWS, Spark ETL, Terraform, Prometheus, Grafana • 50% faster data warehouse ETLs — optimized critical PostgreSQL pipelines moving data into the analytics warehouse through query rewrites and indexing improvements • Monolith decomposed into microservices — proposed service boundaries and led the migration to Java 11, Jersey, and Swagger, improving deployment independence across teams • Observability introduced from scratch — created Prometheus custom metrics, Grafana dashboards, and alerting rules where none previously existed, giving the team first-time visibility into service health

Tech Stack: Java 11, Spring Boot, PostgreSQL, Kafka, Elasticsearch, Redis, AWS, Spark ETL, Terraform, Prometheus, Grafana • 50% faster data warehouse ETLs — optimized critical PostgreSQL pipelines moving data into the analytics warehouse through query rewrites and indexing improvements • Monolith decomposed into microservices — proposed service boundaries and led the migration to Java 11, Jersey, and Swagger, improving deployment independence across teams • Observability introduced from scratch — created Prometheus custom metrics, Grafana dashboards, and alerting rules where none previously existed, giving the team first-time visibility into service health

Tech Stack: Java, Kotlin, Spring Boot, Oracle SQL, Redis, REST, SOAP, WebLogic • Legacy invoice parsing system replaced — designed and built a barcode parsing microservice handling ~3,000 calls/day to parse invoices from multiple external providers, replacing a fragile legacy system with a maintainable Spring Boot service • Platform modernized — migrated the Exchange Rate application from Java 7 to Java 8, upgraded framework dependencies, and delivered new product features with no disruption to existing functionality

• Greenfield internal platform delivered, built an internal staff management platform using Spring Boot, MySQL, N-tier architecture, JUnit 5, Mockito, and Kotlin for service-layer logic