Andrea Matté

Backend engineer on Smartpricing, the AI-powered RMS inside the Smartness platform (4,000+ properties in Europe).

• Rewrote the core pricing engine that produces all dynamic prices for 4,000+ hotels; migrated gradually with no downtime. Write-up: From custom Python per customer to a parameterized engine.
• Built from scratch an event scraping and deduplication pipeline, using connected-components graph algorithms and distance-based geospatial filtering to keep stable identifiers across runs. Write-up: Designing a convergent event deduplication pipeline.
• Co-developed the competitor price scraper, its APIs, and a ScyllaDB storage layer holding ~300 GB compressed at a ~100x compression ratio.
• Led multiple legacy-to-modern migrations of backends and infrastructure with zero-downtime cutovers.

ScyllaDB storage for scraped competitor prices, optimized for high compression and fast reads.

• Keeps the full dataset online in ~300 GB, with ~100x shrink vs raw JSON and ~20x logical-to-physical compression in production, while serving ~100 QPS and cutting storage by 90% vs the previous format.
• Designed the partition and clustering key layout for horizontal scale and compression locality, with a custom binary vector encoding for prices.
• Built the ingestion path from the scraper into Scylla via batching, Kafka and stream processing, keeping write amplification low and absorbing bursts.
• Technical deep dive: Achieving ~100x Compression on Scraped Pricing Data in ScyllaDB.