Autoplius.lt — The Largest Automotive Classifieds Portal in the Baltic States

Users need fast and relevant search across more than 210,000 listings with numerous filters (make, model, year, mileage, price, region, body type, etc.). Heavy SQL queries could not handle the load and did not deliver the required relevance.

• Designed and optimized ElasticSearch indices for key entities (listings, cars, dealers, spare parts).
• Implemented data synchronization between the DB and ElasticSearch via RabbitMQ events.
• Configured faceted search, aggregations, and sorting for filters.
• Optimized mappings and reindexing for large volumes.

Publishing, editing, moderating, and indexing listings are heavy operations that cannot be performed synchronously in the HTTP cycle. Under growing load, this caused delays and timeouts.

• Designed event queues for publishing, updating, deleting, and moderating listings.
• Implemented workers using Symfony Messenger + RabbitMQ.
• Added retry policies, a dead-letter queue, and monitoring.
• Ensured idempotency of handlers.

Popular pages (homepage, categories, top listings, dealer pages) generated enormous DB load on every request. Faster delivery and reduced load were required.

• Introduced multi-level caching of hot data and aggregates in Redis.
• Configured cache invalidation via events (listing publish/edit/delete).
• Used Redis for sessions and distributed locks.
• Added monitoring of hit/miss and response times.

Buyers need access to vehicle history reports directly on the listing page. A reliable integration with the external Autoistorija.lt service was required without blocking user requests.

• Implemented an integration client with Autoistorija.lt with response caching.
• Moved heavy requests into asynchronous tasks via RabbitMQ.
• Added error handling, timeouts, and fallback scenarios.
• Ensured report rendering in Twig templates on the listing page.

Car dealers need tools for market assessment, analysis of listing performance and competitive landscape, and automation of listing management. Existing capabilities were insufficient for assortment planning.

• Developed the backend for Autopulsas: market data aggregation, segment metric calculation.
• Implemented analytics on dealer and competitor listings.
• Added bulk listing management tools (bulk operations).
• Built reports and exports for dealers.

Private sellers need simple tools: car valuation and the ability to pay for a listing after the car is sold ("Mokėk, kai parduosi"). Correct handling of payments and service terms was required.

• Implemented a car valuation service based on market data.
• Developed the "Mokėk, kai parduosi" mechanics: deferred payment, sale tracking, charge per terms.
• Integrated the payment gateway and status processing via RabbitMQ.
• Added notifications and a seller dashboard.

The portal serves nearly 600,000 users and over 11 million visits per month. Stable operation under peak load without response degradation was required.

• Profiled bottlenecks and optimized SQL queries and indices.
• Introduced caching and asynchronous processing of heavy operations.
• Optimized Twig templates and static asset delivery.
• Configured horizontal scaling of workers.

The portal operates 24/7, and failures in indexing, queues, or integrations are business-critical. A monitoring and alerting system was required.

• Set up collection of metrics for ElasticSearch, RabbitMQ, Redis, and key business events.
• Implemented alerts for anomalies (DLQ growth, indexing delays, worker errors).
• Added dashboards for the team.
• Integrated logging of critical operations.

Hundreds of thousands of listings are published on the portal, including duplicates, fraudulent, and incorrect posts. Automated moderation was required without delays for legitimate sellers.

• Implemented a moderation pipeline: automated rules + manual review queue.
• Added checks for duplicates and suspicious patterns.
• Moved moderation into asynchronous tasks via RabbitMQ.
• Added tools for moderators and decision logging.

All actions on listings (creation, editing, moderation, deletion) needed to be tracked for dispute resolution and analytics. An audit system was required that does not slow down core processes.

• Designed an audit log schema linked to listing, user, and action.
• Implemented asynchronous audit writing via RabbitMQ.
• Added tools for searching and exporting audit logs (ElasticSearch).
• Ensured immutability of records.

Growing audience and number of listings required infrastructure scaling without rewriting key components. A solution resilient to peak loads and data growth was required.

• Moved heavy operations into separate horizontally scalable workers.
• Configured ElasticSearch index sharding.
• Optimized Redis and RabbitMQ usage.
• Introduced load testing and profiling.