Senior DevOps Engineer - Remote

Replika

2d•Remote

About The Position

We need a senior DevOps engineer who treats developer experience as a first-class product. You’ll work alongside our existing DevOps engineer, who’s focused on migrating legacy services into a clean deployment story. Your job is different: make every engineer at Replika ship faster, with less friction, fewer footguns, and more confidence. This isn't a hyperscaler role - we serve millions, not billions, so the work looks more like smart trade-offs across a small senior team than mega-scale architecture. We need someone who understands how engineering actually works, writes real code, and genuinely cares about the daily experience of the people building the product. You'll also need product fluency — not at a backend engineer's depth, but enough to know what runs on which cluster and why. If image moderation can run as a batch job on cheaper preemptible GPUs with a few seconds of latency, you should spot the cost optimization and propose it, not wait for someone else to.

Requirements

5+ years in DevOps, SRE, or platform engineering at companies where you actually owned developer experience, not just kept the lights on.
Strong coding skills. You write production-quality Python (or similar), not just shell scripts and YAML. You can read backend code well enough to debug a deploy issue without escalating.
Deep experience with Kubernetes and Docker. You can debug a pod that’s misbehaving, design sensible deployment configs, and reason about resource limits without guessing.
Hands-on with cloud platforms (AWS or GCP). You’ve managed real workloads in production, not just done a certification.
CI/CD chops. TeamCity experience is a plus, but more important is that you’ve owned a pipeline end-to-end and made it fast, reliable, and easy to understand. We use TeamCity and GitHub Actions.
You think in terms of developer experience. You’ve made other engineers’ lives meaningfully better and can talk specifically about how.
Product instincts. You’re comfortable asking “what does this service actually do” and using the answer to make infra decisions.
Strong English (B2+). We’re distributed across time zones — clear written communication matters.
Familiarity with observability stacks (Prometheus, Grafana, Datadog, or similar) and a view on what’s worth alerting on versus what’s noise. You find observability gaps and make sure they are filled.

Nice To Haves

Experience with GPU clusters and ML workloads — model serving, autoscaling inference, cost optimization for GPU-heavy services.
Interest, backed with projects, in AI-first engineering workflows for AI assisted engineering.
Familiarity with observability stacks (Prometheus, Grafana, Datadog, or similar) and a view on what’s worth alerting on versus what’s noise.
Experience with Terraform, Pulumi, or another IaC tool. You version your infrastructure.
You’ve introduced a developer experience improvement at a previous company that you’re genuinely proud of — something you can describe in detail and quantify the impact of.
Experience at a B2C company with real users on the line, where downtime translates directly to user trust.

Responsibilities

Own developer experience end-to-end: deployments, staging, dev environments, CI/CD pipelines. Find what’s painful, fix it, then make it stay fixed.
Be a force multiplier for the engineering team. Every hour you save the team compounds. You think in those terms.
Maintain, improve, and sometimes rebuild our CI/CD pipelines. Deploys should be safe by default — it should be hard to break things and easy to recover when something does break.
Work alongside our existing DevOps engineer on the broader infra picture: cluster management, observability, secrets, networking, cost. You split the work based on what each of you is best placed to own.
Build and maintain templates and tooling for spinning up new services and dev environments. The 47th time someone needs a new service, it should take 10 minutes, not 3 days.
Understand what runs on our clusters and why. Recommend and implement efficiency wins where you see them — right-sizing, autoscaling, scheduling, GPU utilization. You don’t need permission to make sensible cost calls.
Write real code when the situation calls for it. Internal tooling, glue services, automation scripts — you don’t hand it off to a backend engineer.
Care about the small stuff: clear runbooks, sane defaults, good error messages, deploys that fail safely. The kind of work that nobody notices when it’s working well — which is the point.