Rita Lopes

I started my career staring at SIP traces at 2am, wondering why calls were dropping. Years inside telecommunications and real-time infrastructure taught me that the hard part is rarely the query. It is the storage decision you made eighteen months earlier and cannot undo. I now run Gigapipe, where we build observability tooling for people drowning in telemetry. Our open-source backend has been ClickHouse-backed since day one and has passed 500,000 container pulls, so I have spent real years on sharding, replication and rebalancing under load. Lately I have been building a storage engine from scratch in C++ as a DuckDB extension, published open source as RawDuck. Schema-less ingestion, typed columns derived from the data, and a hot and cold tiering layer that moves rows into a shared lakehouse in the background. This talk comes out of that work. Specifically, out of the week I spent running 50 concurrent processes against a Postgres-backed DuckLake catalog and killing them at the worst possible moment.