Staff Engineer

This role is responsible for two things that are more connected than they first appear: (1) making production deployments safe enough that rollback is the default response to a problem, and (2) making configuration changes safe, audited, and AI-supervised so that bad configuration is caught before it reaches production. Both are fundamentally release-safety problems.

GreyOrange operates a platform that multiple enterprise customers (DHL, XPO, Continental, others) run in production warehouses. A bad deployment or a bad configuration change is a customer-visible incident. The infrastructure you build is what stands between a code change and a customer outage. The company's ability to keep all customers on the same major platform version — rather than maintaining separate branches per customer — depends entirely on whether rollback is fast, trustworthy, and routine.

After establishing the deployment and configuration management foundations, you expand into data and messaging engineering — specifically Kafka schema contracts and the Redis claim-check pattern for payload management. The overlap between distributed delivery systems (Kubernetes + Kafka both require schema and contract thinking) makes this a natural progression.

What You Will Do

First Six Months - Deployment Safety & Configuration

• Take ownership of the deploy and release engineering domain. Audit the existing Spinnaker pipelines and ArgoCD application definitions to understand the starting point.
• Conduct a Consul configuration inventory — catalog every human-edited key, identify overlaps with the existing configuration management system, and publish a written boundary document clarifying which configs live where.
• Design and ship blue-green progressive delivery for the top three critical stateless services using ArgoCD Rollouts. Write a shared Rollout template that all subsequent services follow.
• Standardise Spinnaker pipeline templates for VM-deployed services. Define smoke-test requirements as a mandatory promotion gate.
• Ship the Java/Spring feature-flag SDK, piloted against one service. Then Erlang and React SDKs.
• Implement Consul-as-backend: bring human-edited Consul keys behind the configuration management UI. Remove direct write access for humans.
  
  

Next Six Months - Configuration Intelligence & Messaging

• Ship AI-supervised configuration: train or fine-tune a model on historical incident data, surface risk warnings in the configuration UI. First version can be rule-based with an LLM layer on top.
• Wire the configuration audit log into the compliance evidence pipeline so configuration changes become part of the joined audit trail alongside access events and incident records.
• Own Kafka schema contract rollout: Schema Registry adoption and Avro migration for the highest-traffic topics. Partner with service teams through the migration.
• Build and ship the Redis claim-check SDK. Monitor payload-size distribution and consumer lag. Demonstrate measurable reduction in Kafka payload sizes on the pilot topics.
  
  

Ongoing

• Run rollback drills on a regular cadence — each customer environment should execute at least one controlled rollback per half so the process is practised before it is needed under pressure.
• Maintain the deployment and configuration systems as production services — they are part of the platform team's on-call rotation.
• Contribute to cross-team architecture reviews with a deployment safety and release-engineering lens.
• Write and maintain design documents and runbooks for everything you build.
  
  

Requirements

• 8+ years of backend or platform engineering with strong Kubernetes and GitOps depth
• Production experience with ArgoCD, ArgoCD Rollouts, Flagger, or comparable progressive-delivery tooling — has configured traffic shifting, canary analysis, and automated rollback in a real production environment
• Has built or substantially extended a feature-flag or runtime configuration management system — not just used a commercial tool, but understood or built the machinery underneath
• Familiarity with Consul, Vault, or the broader HashiCorp ecosystem at an engineering level
• Experience with Spinnaker or a comparable continuous delivery platform for VM-based deployments
• Uses deployment frequency, change failure rate, and mean time to restore as engineering metrics — not just awareness of what DORA is, but a track record of using it to prioritise engineering investment
• Comfortable with heterogeneous deployment environments - both Kubernetes and traditional VMs as first-class production targets in the same organisation
  
  

Nice to Have

• Experience with Apache Camel or Apache Flink - relevant for future work in integration and streaming engineering
• Background with Kafka Connect, schema registries, or MQTT integration - the platform bridges Kafka and MQTT for robot communication
• Multi-tenant SaaS background where customer-segmented feature rollouts are standard practice - GreyOrange deploys to multiple enterprise customers with different risk tolerances and release windows
• Familiarity with LLM APIs (Vertex AI, OpenAI) and retrieval-augmented generation - the AI configuration risk-detection work uses an LLM layer on top of historical incident data
• Erlang/OTP experience or genuine curiosity about it - the Erlang feature-flag SDK requires understanding how OTP applications load and manage configuration at runtime