Distributed Systems Architect: Bare-Metal and High-Scale
Analog · Abu Dhabi · onsite
Analog builds industrial intelligence infrastructure for the physical world: high-throughput data pipelines, real-time analytics, and edge-to-cloud connectivity for clients who can't depend on the cloud. Think regulated environments, air-gapped sites, physically demanding settings where there's no AWS console to fall back on when something breaks.
That's where you come in. Most distributed systems engineers have only ever worked on top of a managed cloud service. This role is different: you'll design the actual infrastructure underneath, the clustering, replication, and failover that keeps messaging, streaming, and databases alive when a client's hardware is sitting in a locked room with no internet connection. Your architecture is what stands between a client and data loss at petabyte scale.
WHAT WE'RE LOOKING FOR, AND WHY
We need someone who's built the thing other engineers just rent. Not someone who's deployed a managed Kafka instance, someone who's clustered it themselves, rack by rack, and knows exactly what happens when a node drops mid-write. Analog's clients don't get to call AWS support, so the person designing their infrastructure has to have already lived through the failure modes that cloud providers hide from everyone else.
WHAT YOU'LL BE WORKING ON
- Mapping Analog's cloud components to their open-source, on-prem equivalents (Kafka/Redpanda, EMQX, ClickHouse/Druid, Ceph/MinIO), and documenting exactly what's gained and lost in each swap
- Designing multi-node clustering, rack-aware replication, and quorum topologies for bare-metal Kafka/Redpanda and Postgres (Patroni) clusters
- Tuning Linux at the level most people never touch: kernel parameters, NUMA bindings, NVMe I/O scheduling, CPU affinity, IRQ balancing, all to eliminate hardware bottlenecks under petabyte-scale write loads
- Writing implementation-ready blueprints and runbooks precise enough that our DevOps team can execute them with Terraform and Ansible, no architectural guesswork required
- Reviewing client hardware specs before deployment and giving a clear go or no-go call
WHAT MAKES THIS ROLE DIFFERENT
There's no layer of management between your design and what actually ships, if your blueprint is wrong, you'll know, because it's the DevOps team that automates exactly what you hand them. This is implementation-grade architecture, not a slide deck: the quality of your output is the difference between a client's system surviving a hardware failure or not. You'll also be building a reusable library of reference architectures that every future air-gapped deployment draws from, so the work compounds.
SKILLS WE NEED
- 10+ years architecting and operating distributed infrastructure at TB to PB scale, in production
- Expert-level Kafka or Redpanda clustering: partitioning, ISR tuning, consumer groups, compaction, full operational lifecycle
- Deep Linux systems engineering: kernel networking (TCP buffer tuning, interrupt coalescing), storage fabrics, NUMA-aware process binding
- A real track record running HA database clusters without a cloud load balancer (Patroni, Pacemaker, or equivalent) in production
- Experience deploying Ceph or MinIO on physical hardware, including CRUSH maps, erasure coding, and performance tuning
- The ability to write blueprints and runbooks that another team can execute without needing you in the room to interpret them
WHO SHOULDN'T APPLY
If your distributed systems experience is entirely on managed cloud services, this isn't the role to learn bare metal for the first time. Same if you need a layer of process or a large team around you to make architectural calls, or if what you actually want is to oversee infrastructure rather than build and tune it yourself. This is a hands-on role for someone who wants to be the one person a client's survival depends on.