Blog
Technical write-ups on backend engineering, data systems, and performance work.
-
Designing a Convergent Event Deduplication Pipeline
How to merge multi-source event records with graph-based deduplication, aggressive search-space reduction, and stable IDs across runs.
-
Achieving ~100x Compression on Scraped Pricing Data in ScyllaDB
How we designed a storage layer for competitor hotel prices that compresses ~30 TB of raw scraped data into ~300 GB in ScyllaDB, with ~20x logical-to-physical compression in production.
-
From Custom Python per Customer to a Parameterized Pricing Engine
How we moved from per-customer pipelines (startup-speed trade-offs) to one parameterized engine, migrated gradually with no downtime, and shipped previews before saving changes.