Staff Software Engineer, Dune

• You are a backend engineer who has gone deep on data systems, or a data engineer who became a strong software engineer. You ship production services, not only pipelines
• You have built or materially extended orchestration and scheduling systems, and can explain precisely what breaks at scale and why
• You have handled schema evolution and data correctness in a system with real consumers downstream, where a breaking change has a cost
• You have built or operated stateful stream processing in production (Flink,Kafka Streams, Spark Structured Streaming, RisingWave, Materialize, Feldera)
• You have strong SQL and modeling skills on large datasets, and an interest in how the query engine underneath actually executes your work
• You have solid computer science fundamentals and distributed systems understanding
• You debug independently and drive root cause analysis to a fix that holds
• You use AI tools well enough that they have changed how you work, you understand their failure modes and dislike ai-slop
• You communicate clearly in writing and get the best out of a distributed team
• If you’re excited about this role but your experience doesn’t align perfectly with every qualification in the job description, we encourage you to apply anyway. You may be just the right candidate for this or other roles

Desirable

• Data Products builds and owns datasets end to end: from raw chain data through decoding to the 3000+ models and 4 petabytes we curate, share directly with customers, and replicate into their warehouses
• The role will focus on the lifecycle of building high quality data: orchestrating thousands of interdependent models, propagating schema changes without breaking downstream consumers, propagating corrections
• That is a software architecture problem in a data domain. This role is a hybrid: a backend engineer who thinks in systems and contracts, working on data
• You will be the engineer we hand ambiguous product requirements to, and will come back with a design, a sequence, and work the team can pick up, while building the hardest parts yourself
• Design and build the control plane for our curated data lifecycle: dependency-aware orchestration, backfills, restatements, retries, partial failure, and recovery
• Decide, dataset by dataset, whether the answer is a model, a service or a job, and own that architecture through production
• Design the contracts between ingestion and curation so a dataset can be reasoned about end to end
• Build alerting and data quality signals that catch real problems and stay quiet otherwise, so on-call is about incidents rather than noise
• Work across Go, Kotlin, Rust, Python and SQL, choosing the right tool rather than the familiar one
• Break large problems into work other engineers can own, and sequence it so we ship something useful early