Tudor is a Senior Software Developer with 10 years of web development experience as a technical contributor on web apps and API projects using Python and applying TDD best practices to deliver high-quality software solutions. With a focus on back-end development using Python, Django, and Flask, he has worked on small and large complex systems, evolving them to be resilient and highly scalable by applying DDD principles and event sourcing. In the last two years, Tudor has been involved in adopting and developing product features powered by stream technologies like Apache Flink and Apache Kafka. He employs a client-centered, supportive, and collaborative approach to iterating and delivering solutions and products for clients.
Implemented claim check pattern inside Celery by offloading large messages to GCS, which reduced memory usage on RabbitMQ by ~30% and improved its stability.
Optimized a couple of systems that powered real-time features, dropping the end-to-end latency from P99 2.5s to P99 250ms, allowing Phone team to build more features on top of the same systems.
Led the transition from Google Pub/Sub to Apache Kafka, now serving as the foundation of the streaming platform.
Contributed to the shift from Apache Beam to Apache Flink and currently driving the Apache Flink platform's adoption across the company.
Overseeing the successful adoption by more than three teams, previously unfamiliar with Apache Flink, in developing and deploying pipelines in production.
Platformized Debezium, including performance improvements contributed upstream, enabling any team at Gorgias to stream data from any Kafka topic.
Developed a Python library to support the transition to an event-driven architecture, enabling easy definition and writing of AVRO serialized events into an outbox table for all Python-based services.
Ported a SQL-heavy feature from PostgreSQL to Elasticsearch, reducing the endpoint error rate from 10% to 0, enabling the display of more accurate statistics for customers and decreasing Redis costs by 20%.
Played a key role in designing, implementing, and deploying a Change Data Capture architecture that involved Debezium, Google Pub/Sub, and Apache Beam to create a denormalized view of multiple PostgreSQL tables as search indices on Elasticsearch.
Successfully reduced by 300x the p99 response time.
Removed 2TB of indices from the PostgreSQL database, resulting in better vacuum performance.
Implemented new features on existing services, cleaning up Ansible playbooks and splitting them into roles and handling Nomad deployments.
Engineered a mechanism to handle database downtimes on the service and developed a new solution to store audit logs on the database using a custom logging adapter.
Offered solutions for complex technical issues, improving the performance of a few SQL queries by adding missing table indexes and fixing queries on the project.
