Principal Software Engineer - Object Store

Roblox's Object Store team is building the in-house, multi-tenant object storage platform that underpins every large-object workload at Roblox — assets, media, analytics, backups, and machine learning datasets — at exabyte scale, trillions of objects, and terabytes per second of throughput. We build and run it ourselves on bare metal in our own data centers, replacing cloud object storage with a platform we control end to end — down to the hardware. As a principal engineer on the Object Store team, you will shape the architecture, build and launch the capabilities that keep this platform durable, fast, and efficient at global scale, and set the technical direction for where it goes next. You will report to the Engineering Manager for Storage.

You will:

• Design and build the object store — S3-compatible gateway, metadata and indexing, data placement and tiering, multi-tenancy and isolation.
• Own the technical strategy: multi-year architecture, build-vs-buy calls, and the migration off cloud object storage.
• Evolve the control plane for elastic scaling, autonomous healing, self-service provisioning, and zero-downtime tenant moves.
• Run the bare-metal fleet at scale — cluster lifecycle, upgrades, rebalancing, failure domains, hardware and capacity planning.
• Profile and optimize critical data paths; drive down tail latency and cost per byte.
• Establish engineering best practices — design reviews, benchmarks, failure drills, post-incident retrospectives — targeting 4+ 9s.
• Automate everything: testing, CI/CD, rollout safety, observability, autoscaling.
• Mentor engineers and partner cross-functionally with security, analytics, and product teams.

You have:

• 8+ years of experience in software engineering or a relevant field
• Deep experience building and operating large-scale distributed storage — object stores, blob stores, or distributed file systems — at petabyte scale or beyond.
• Hands-on experience running an object storage platform in production: data placement and rebalancing, erasure coding, cluster tuning, and upgrades at scale. Depth operating these systems matters more here than contributing to their internals.
• Strong skills in C++, Go, or Rust.
• Deep understanding of durability and replication — erasure coding, geo-replication, consistency trade-offs, blast-radius containment.
• Proven success shipping high-throughput, low-latency services on bare-metal, on-prem fleets with strong observability.
• Ability to turn ambiguous requirements into a clear roadmap and influence cross-functional stakeholders.
• Passion for automation, rigorous testing, and data-driven decisions.
• Bonus: upstream contributions to open-source storage projects; S3 API compatibility at scale, NVMe/HDD tiering, or migrating large workloads off public-cloud object storage.