Principal Database Infrastructure Engineer, VideoAmp

• 8+ years of software engineering experience with significant depth in database infrastructure, distributed systems, or data platform engineering
• Strong systems programming in Rust, or deep C++ or Go experience with a clear path to Rust, including async runtimes such as Tokio and concurrent data structures
• Proven experience building or significantly modifying a distributed data system such as a query engine, stream processor, distributed database, or large-scale data pipeline, with a solid understanding of shuffles, partitioning, and network and memory bottlenecks
• Fluency with columnar formats and vectorized execution, including Arrow, Parquet, and the mechanics behind their performance characteristics
• Strong grounding in distributed systems fundamentals: consistent hashing, leader and heartbeat protocols, backpressure, partial failure, and graceful degradation
• A performance engineering mindset: you profile before optimizing and defend changes with real benchmark numbers

Desirable

• The Principal Database Infrastructure Engineer will serve as a technical cornerstone of VideoAmp's Database Infrastructure team, driving the design and execution of scalable, production-critical data systems that power VideoAmp's platform
• This is a high-impact individual contributor role at the intersection of distributed database engineering, query performance, storage architecture, and developer enablement
• You will architect and own the foundational database systems serving live customers and internal teams, operate in a rigorous, reliability-driven culture, and help VideoAmp scale its data infrastructure as the platform grows
• Distributed Query Execution
• Design and implement the physical plan distribution pass, including network shuffle, coalesce, and partition isolator insertion. Own the plan serialization codec that ships sub-plans to workers, and maintain the S3 and Flight result exchange paths
• Worker Coordination at Scale
• Own worker discovery and heartbeating, and lead development of the next generation of load balancing: a work-stealing protocol and a replication-aware hash ring, both currently in design, to keep workers evenly loaded and resilient to node loss
• Query Optimizer Internals
• Work across cost-based join reordering, cross-stage bloom filter cascade, scan deduplication, and selectivity estimation. Several of these live in our DataFusion fork; you will upstream where it makes sense and maintain the delta where it does not
• Storage and Caching
• Own the NVMe LRU cache over S3, Parquet read strategies including full-file and range reads, and Iceberg partition pruning and snapshot handling
• Performance Engineering
• Close the remaining gap on queries where we still trail Snowflake, specifically multi-shuffle plans and redistribution after scalar-subquery extraction. TPC-H benchmarks are the scorecard