Platform engineer backend infrastructure

Focal Systems is the industry leader in retail AI solutions. We are a Silicon Valley-based startup headquartered in San Francisco, with operations in Canada, the UK, multiple EU countries and a tech hub in Poland. We are a Deep Learning-first company. Our mission is to automate and optimize brick-and-mortar retail using deep learning computer vision. Focal Systems has been deployed at scale with some of the top retailers in the world, processing hundreds of millions of images per day across tens of thousands of stores globally.

• Own the design, operation and cost of Focal's production platform on GCP: Kubernetes, Kafka, MySQL, MongoDB, Redis, object storage.
• Drive measurable reductions in COGS and latency, for example lowering GPU inference cost per store through machine types, batching, autoscaling or scheduling, and report the numbers.
• Read, debug and change Python (FastAPI, Kafka consumers) backend services. Platform problems are often fixed in application code, for example a MySQL query that ran fine at 1,000 stores and times out at 10,000.
• Fix throughput bottlenecks such as consumer lag at store opening, whether the cause is partition count, consumer throughput, a downstream database or autoscaling reaction time.
• Rework metrics, alerts and dashboards so the on-call engineer can tell which stores and regions are affected when an alert fires. Take part in incident response, postmortems and capacity planning.
• Build tooling that gives engineers access to data across regions and storage systems in minutes instead of days.
• Maintain infrastructure-as-code (Terraform, Helm) and the deployment path to production, and work with the backend, deep learning and edge teams so the platform keeps up with new products and models.

• Python as a software engineer, not only for scripting. You will contribute to backend code every week.
• 5+ years building and operating production backend systems on a cloud platform. GCP preferred; deep AWS or Azure experience is fine.
• Hands-on production experience operating at least one stateful system under load: Kafka or another streaming platform, MySQL/PostgreSQL, or MongoDB. We mean partitioning, replication, lag, slow queries and failover, not only provisioning.
• Kubernetes in production, including scheduling, resource limits, autoscaling with KEDA, GPU node pools and debugging misbehaving pods.
• Debugging production incidents across layers (application, messaging, database, orchestration, network), with examples you can walk us through.
• Observability with Prometheus, Grafana, Grafana Mimir and Sentry, or equivalent: instrumenting code, writing useful alerts.
• Terraform or another infrastructure-as-code tool.
• You take ownership of problems end to end and follow them into whichever layer they lead.
• You make decisions from data (profiles, metrics, query plans, cost reports) and prioritize well in startup/scaleup environments.
• Advanced written and spoken English.
• Comfortable coordinating across time zones with a global team.
• Comfortable with a shared on-call rotation.
• Будет плюсом: MySQL query tuning; Kafka internals (consumer group rebalancing, partition strategy, throughput tuning), Confluent Cloud and Kafka Connect; GPU workloads or ML inference serving in production (Ray/KubeRay or similar), and cloud cost optimization with measured results; Go, Snowflake, Airflow, CircleCI, Helm, and blue/green or canary releases

• Competitive Salary & Attractive Stock
• Paid Time Off
• Team Retreats
• Education grants