# Andy Pavlo Joins ClickHouse Labs: Database Research Shifts > Source: > Published: 2026-08-04 00:12:47+00:00 Andy Pavlo — CMU database professor, VLDB Early Career Award winner, and founder of OtterTune — [announced this week](https://clickhouse.com/blog/andy-pavlo-joins-clickhouse) he’s joining ClickHouse as VP of Database Research to found ClickHouse Labs. The announcement dropped August 3, 2026, and it’s one of the more significant personnel moves in the database world in recent years. A tenured CS professor with this pedigree doesn’t leave academia for a startup on a whim. Pavlo has led Carnegie Mellon’s Database Research Group since 2013. His work on autonomous and self-driving databases — systems that tune themselves without human intervention — earned him the ACM SIGMOD Jim Gray Dissertation Award, a Sloan Research Fellowship, and the NSF CAREER Award. He co-founded OtterTune in 2020, an ML-driven database configuration tuning service. The man knows what a database looks like from the inside, at research depth. Now he’s taking that expertise directly into ClickHouse’s production environment. ## What ClickHouse Labs Is Building ClickHouse Labs isn’t structured as a traditional research division isolated from the engineering team. Pavlo was clear about this in the [founding post](https://clickhouse.com/blog/andy-pavlo-founding-clickhouse-labs): “The most exciting database research does more than describe a new idea in a paper. It demonstrates what is possible by building and testing real systems.” The lab will work directly alongside engineers and validate findings against ClickHouse’s live production traffic. The research agenda covers four areas: database and data architectures, query processing, systems performance, and infrastructure for both ClickHouse and PostgreSQL. The AI framing matters here. ClickHouse acquired Langfuse — an LLM observability startup — earlier this year as part of its $400M Series D, signaling that AI workloads are a first-class priority. Pavlo’s lab will specifically address how databases need to evolve when AI agents are the primary consumers: higher query volumes, tighter latency requirements, and access patterns no traditional DBMS was designed for. His specialty — autonomous database optimization — has an obvious application here. OtterTune proved that ML can tune database configurations better than human DBAs. ClickHouse now has millions of queries in production to train against. Expect ClickHouse Cloud to eventually self-tune query plans and auto-optimize configurations at a scale no academic prototype ever touched. Related:[Turso Is Building Postgres in Rust — What That Means for Your Database Stack] ## Industry Is Where Database Research Lives Now Pavlo’s move is another data point in a longer trend. The most consequential database research in the last two decades didn’t happen in universities — it happened at companies with production traffic to validate against. Google published MapReduce, Bigtable, and Spanner from internal systems handling real-world scale. DuckDB came out of CWI, a research institute with direct ties to industry. Snowflake was founded by database researchers who left Oracle to build at startup speed. Academic research has a fundamental constraint: experiments run on synthetic datasets or TPC-H benchmarks. Pavlo himself identified this problem. ClickHouse Labs will run experiments on workloads processing trillions of rows, from thousands of production customers. That’s a fundamentally different research environment — and one where results are more trustworthy. When a new query optimization trick passes muster on ClickHouse’s real traffic, it actually means something. The parallel to IBM Research in its prime isn’t far-fetched. IBM Research invented relational databases, co-developed RISC architecture, and gave the world SQL itself. That era of industry-based foundational research produced more lasting CS contributions than many decades of academic work. ClickHouse Labs isn’t going to reinvent SQL, but the ambition to do production-validated foundational research at this scale is a meaningful signal about where the field is heading. ## Why ClickHouse Made the Right Bet ClickHouse’s position makes the investment make sense. After its [$400M Series D in January 2026](https://www.businesswire.com/news/home/20260803890510/en/) — at a $15B valuation with 250% ARR growth year-over-year — the company has both the resources and the urgency to invest in research depth, not just engineering speed. It’s the dominant open-source OLAP database, competing with Snowflake and BigQuery for AI-driven analytics workloads where raw query performance alone won’t be enough to differentiate. Pavlo cited three reasons he chose ClickHouse specifically: “exceptional engineering culture, a strong open-source community, and workloads that push database technology to its limits.” That last part is the key. ClickHouse’s vectorized execution engine, built in C++ with extensive SIMD optimization, is exactly the kind of system where rigorous research compounds. Pavlo said he tracked ClickHouse since its 2016 open-source launch and was initially skeptical — before the engineering quality won him over. ## Key Takeaways - Andy Pavlo — CMU database professor and OtterTune founder — is joining ClickHouse as VP of Database Research to lead ClickHouse Labs, announced August 3, 2026. - ClickHouse Labs is integrated directly with engineering, not isolated — research will be validated against production traffic from thousands of customers, not synthetic benchmarks. - The research agenda is AI-first: autonomous query optimization, databases built for agentic workloads, and foundational improvements for both ClickHouse and PostgreSQL. - Pavlo’s self-driving database work (OtterTune, NoisePage) maps directly to ClickHouse at scale — autonomous configuration tuning in ClickHouse Cloud is the logical next step. - Watch ClickHouse Labs publications. When a researcher with Pavlo’s track record validates ideas on production workloads at this scale, the results matter.