Platform Engineer, Kubernetes
Atmora Tech
- Home
- Careers
- Platform Engineer, Kubernetes
- Jodhpur, India
Atmora runs more than 40 client workloads across EKS and AKS, from a nine-node cluster carrying Kestrel Industrial's shop-floor telemetry to a two-region active/passive setup for Corvus Logistics. You will own the internal platform those delivery teams deploy onto: the Terraform module library, the Argo CD app-of-apps layout, and the paved path that takes a service from merge to production in under 12 minutes.
- Full-time — on site
- 4-7 years in infrastructure, including at least one production Kubernetes estate you did not inherit already working
- Jodhpur, India
The work
What you will do
- Maintain the Terraform module library used by every engagement, and keep drift close to zero through a weekly plan-and-review cycle across all accounts.
- Run progressive delivery with Argo Rollouts — canary at 5 / 25 / 100 per cent with automated rollback on error-rate and latency gates.
- Bring cloud spend down on two named accounts by 20-30 per cent through right-sizing, spot node pools and storage tiering, without moving any SLO.
- Own the observability stack (Prometheus, Loki, Grafana), including a cardinality budget currently set at 1.5 million active series.
- Handle escalations on the infrastructure rotation and run blameless reviews afterwards.
What we look for
- Kubernetes past the tutorial layer: scheduling, admission webhooks, PodDisruptionBudgets, and why your rolling update stalled.
- Terraform written as reusable modules with sensible variable surfaces, not copied directories.
- Linux networking you can debug — iptables or eBPF paths, DNS resolution inside a pod, MTU mismatches on overlay networks.
- Go or Python good enough to build a controller or a one-off migration tool.
- Judgement about when Kubernetes is the wrong answer and a pair of VMs behind a load balancer would serve the client better.
Other openings
- Bengaluru, India
- Full-time — hybrid, three days a week on site
You will work on the settlement and ledger services behind Meridian Exchange's B2B marketplace: roughly 40 million ledger rows a month, Java 21 on Spring Boot, Postgres 16, Kafka for event fan-out. The interesting part is not new endpoints, it is correctness under load — idempotent writes, exactly-once effects on top of at-least-once delivery, and holding p99 write latency under 250 ms as the ledger grows.
- London, UK
- Full-time — London office, with four to six client-site days a month across the UK and EU
Most of our UK work is legacy modernisation for regulated businesses: Cordage Bank moving a 22-year-old core banking periphery off a shared Oracle schema, Aldridge Risk replacing an overnight batch that had grown to 9 hours and 40 minutes against a 6-hour window. You will run discovery, decide the migration shape, and stay accountable for it once delivery starts — not hand over a slide deck and leave.
- Bengaluru, India
- Full-time — hybrid, three days a week on site
Three active workstreams: price-anomaly detection over Northwind Commodities' trade blotter, document extraction for Orbit Supply covering roughly 30,000 pages a month of bills of lading and purchase orders, and retrieval over a contract library of 140,000 documents. Every one of them starts with an evaluation set, not a model. If you cannot measure it against a labelled sample the client agrees with, we do not ship it.
- Home
- Careers
- Platform Engineer, Kubernetes
- Jodhpur, India
Atmora runs more than 40 client workloads across EKS and AKS, from a nine-node cluster carrying Kestrel Industrial's shop-floor telemetry to a two-region active/passive setup for Corvus Logistics. You will own the internal platform those delivery teams deploy onto: the Terraform module library, the Argo CD app-of-apps layout, and the paved path that takes a service from merge to production in under 12 minutes.
- Full-time — on site
- 4-7 years in infrastructure, including at least one production Kubernetes estate you did not inherit already working
- Jodhpur, India
The work
What you will do
- Maintain the Terraform module library used by every engagement, and keep drift close to zero through a weekly plan-and-review cycle across all accounts.
- Run progressive delivery with Argo Rollouts — canary at 5 / 25 / 100 per cent with automated rollback on error-rate and latency gates.
- Bring cloud spend down on two named accounts by 20-30 per cent through right-sizing, spot node pools and storage tiering, without moving any SLO.
- Own the observability stack (Prometheus, Loki, Grafana), including a cardinality budget currently set at 1.5 million active series.
- Handle escalations on the infrastructure rotation and run blameless reviews afterwards.
What we look for
- Kubernetes past the tutorial layer: scheduling, admission webhooks, PodDisruptionBudgets, and why your rolling update stalled.
- Terraform written as reusable modules with sensible variable surfaces, not copied directories.
- Linux networking you can debug — iptables or eBPF paths, DNS resolution inside a pod, MTU mismatches on overlay networks.
- Go or Python good enough to build a controller or a one-off migration tool.
- Judgement about when Kubernetes is the wrong answer and a pair of VMs behind a load balancer would serve the client better.
Other openings
- Bengaluru, India
- Full-time — hybrid, three days a week on site
You will work on the settlement and ledger services behind Meridian Exchange's B2B marketplace: roughly 40 million ledger rows a month, Java 21 on Spring Boot, Postgres 16, Kafka for event fan-out. The interesting part is not new endpoints, it is correctness under load — idempotent writes, exactly-once effects on top of at-least-once delivery, and holding p99 write latency under 250 ms as the ledger grows.
- London, UK
- Full-time — London office, with four to six client-site days a month across the UK and EU
Most of our UK work is legacy modernisation for regulated businesses: Cordage Bank moving a 22-year-old core banking periphery off a shared Oracle schema, Aldridge Risk replacing an overnight batch that had grown to 9 hours and 40 minutes against a 6-hour window. You will run discovery, decide the migration shape, and stay accountable for it once delivery starts — not hand over a slide deck and leave.
- Bengaluru, India
- Full-time — hybrid, three days a week on site
Three active workstreams: price-anomaly detection over Northwind Commodities' trade blotter, document extraction for Orbit Supply covering roughly 30,000 pages a month of bills of lading and purchase orders, and retrieval over a contract library of 140,000 documents. Every one of them starts with an evaluation set, not a model. If you cannot measure it against a labelled sample the client agrees with, we do not ship it.