Backend Engineer (Infra)

About Caliber

Caliber is a people intelligence platform that we deploy custom for each client. A client comes to us with a hiring or talent problem — find the fifty people in the world who could do this job, understand a competitor's org before it moves, build the pipeline for a team that doesn't exist yet. We stand up a deployment for them: their own isolated database of people, companies, jobs and candidates, sitting on top of a shared corpus of 181 million professionals and 141 million jobs, driven by an agent that can search the world, reason over the client's own records, run outreach, and keep everything current. Then we build on that deployment for them — custom screens, custom tables, custom workflows shaped to how they actually hire. What makes this work is the platform underneath: multi-tenant by construction, one corpus serving everyone, one agent tool surface serving every client, and a data layer nobody else has. That platform is what you'd own.

About the role

This is a pure backend and infrastructure role on Caliber core. Not a forward-deployed role, not full-stack. You'd own the engine, the data plane, and the operational reality of a multi-tenant platform running real client workloads — so that the people building client-facing work have something excellent to build on. It's engineer #3, working directly with the founders. Short path from decision to production, and real ownership of the parts that must not break.

What you'll do

Own the runtime. Caliber runs continuous background work — outreach cadences, reply watchers, scheduled agent runs, corpus ingest. Move it onto a durable job queue with the correctness properties it needs: idempotency, at-least-once delivery, safe concurrency across replicas, and no lost work through a deploy. Own the data plane. A 181M-row ClickHouse corpus, a Postgres row store per client, and a compatibility layer between them that must not silently disagree. Ingest pipelines, query performance, cost, and the read traps that produce confidently wrong answers. This is where the moat physically lives. Own deploy and operational safety. Migrations in a pipeline, a staging environment, rollbacks that work, and alerting that tells us a client's outreach stopped before the client does. Multi-tenant means one bad deploy has five audiences. Make the platform multi-engineer. Turn today's per-client extension points into a real contract: what can be built per client, what graduates into core, how client code ships without redeploying everyone, and what stops one deployment's custom lane from taking down another's. You'd write the guardrails and the docs, and be the review that keeps core coherent. Build backend product. This isn't a maintenance role. The cold-email deliverability engine, the campaign event bus, the agent's tool surface, the sourcing funnel over the corpus — all half-built, all backend, all the product.

What we're looking for

- You've operated something real in production — on call, at a company where the outage was yours. ~5+ years is roughly the shape, but we care about what you've been responsible for, not the number.

- Deep Postgres — roles and grants, query plans, migrations against live data, why the schema is the way it is. The most load-bearing skill in the role.

- Distributed systems judgment — queues, idempotency, at-least-once, concurrency across replicas. You've been burned by a rolling deploy and can say exactly how.

- Cloud infrastructure as code, ideally AWS with CDK or Terraform. Comfortable that DNS, secrets and services live in a repo.

- Multi-tenancy taken seriously. You understand why the tenant boundary belongs in the database rather than in a "WHERE" clause somebody has to remember.

- You write for the next person. Caliber is legible because the why got written down every time. That standard doesn't relax when you join it.

Bonus: you've built a platform other engineers shipped on; run analytical stores at 100M+ rows; owned data ingest at scale; done email deliverability or another domain where being wrong costs money you can measure; been the second engineer somewhere and know what the first one always forgets. Probably not a fit if: you need a spec handed to you, you want a large team around you, you're allergic to touching production, you want to be mostly full-stack, you need mentorship, you can't build end to end yourself.

How we work

- Massive Ownership For this role, that means the ability to have the final say on how we build, CI/CD, infra decisions, etc

- Speed Move incredibly quickly. Doesn't mean work 14 hours a day, but means moving fast and biasing toward action

- One production database, and clients sit on top of it The care that this requires is part of the job description.

- In person in NYC, most days. The work is too coupled for remote/async right now.