Why this role exists We run two production databases that drive every dollar of revenue, and we need one owner who treats both as their job, not as a queue of tickets to be chased.
The role exists because of a concrete failure pattern we are eliminating: a server-wide deadlock P0 in our SQL Server reporting procedures was identified, deferred across four planning cycles, and sat unresolved for seven weeks with no escalation and no handover. That cannot happen again. We need someone who owns incidents end to end, communicates proactively, and never lets a P0 go dark. What you own * SingleStore (on-prem, OVH cluster, 900M+ rows): the warehouse behind enrichment, search, and audience preview. Partitioning, sharding and distribution keys, indexing (hash, sorted, full-text), query performance, and the migration of ingestion from AWS Glue to native SingleStore PIPELINES. * SQL Server (customer data, billing, order processing): stored procedures, deadlock and blocking resolution, backup and restore performance, and the data layer feeding order processing. * The shared data layer: as we move DDL into a versioned repo (one source of truth across SingleStore and SQL Server), you are the reviewer and gatekeeper. No more ad-hoc tables created at 7am with no approval. * Performance as a continuous practice, not a fire drill: weekly proactive query and partition reviews, alerting on long-running queries, and a standing optimization backlog that you drive.
The problems you will solve in week one These are real, current, and representative of the work: * A SQL Server reporting procedure deadlocking the server (200+ blocked tasks, scanning 54 tables / 127M records). Diagnose, refactor, and ship the fix with a regression test, not a deferral. * SingleStore partition skew: a single distribution key value (for example a placeholder UCID) collapsing millions of records into one partition. Redesign the distribution and index strategy and validate the rebalance. * The LIKE vs MATCH (full-text) tradeoff on job-title search: queries running 3 to 5 minutes that should run sub-second. Define a defensible correctness metric, then ship the index strategy that hits the target. * Materialized geo aggregates (nightly pre-compute per country and state) to kill cold-start latency on audience-size preview.
Must-haves * 5+ years production DBA / database engineering. Strong SQL fundamentals are the core requirement, plus hands-on ownership of (a) a large-scale distributed or columnar data warehouse (SingleStore, ClickHouse, Snowflake, BigQuery, Redshift, Vertica, Greenplum, or equivalent) and (b) a transactional RDBMS, ideally SQL Server but PostgreSQL, MySQL, or Oracle at scale is acceptable if you can pick up T-SQL fast. You need both the analytical-warehouse and transactional sides; the specific engines are learnable, the instincts are not. Comfort with execution plans, deadlock and blocking analysis, and distributed-system query tuning. * A track record of resolving production P0 and P1 database incidents, with the judgment to escalate early and the discipline to document and hand off. * Proven query optimization at scale (100M+ row tables): indexing strategy, partitioning and sharding, materialized views, query rewriting. * Strong written English and proactive communication. You post status before you are asked. You flag risk before it becomes an incident.
The Claude Code requirement (non-negotiable) You must be a fluent, daily Claude Code user who can independently find and ship optimizations without round-tripping every decision through the CTO.
Concretely, the bar is what we already do internally: pulling real customer query logs, generating a representative test set, running multiple optimization pipelines, and cutting query time 40 to 45% in an afternoon, with the analysis and the fix produced by Claude under your direction. You are expected to operate at that level autonomously: investigate, benchmark, propose, implement, validate. Claude is your leverage, not a toy. If you currently use AI as autocomplete rather than as an engineering partner, this is not the right role. Nice-to-haves * Experience with SingleStore PIPELINES and migrating off ETL tools (AWS Glue). * Familiarity with .NET (the SQL Server side) and TypeScript / GCP (the V2 stack we are migrating order processing and enrichment toward). * Database CI/CD: schema-as-code, migrations, automated test scripts against staging. * Exposure to high-volume B2B data and enrichment pipelines.
