Platform Reliability / DevOps Engineer

Contract engagement · Remote
Engagement:
Independent contractor (30–40 hrs/wk)
Remote from the Philippines
Core hours 6am–2pm Manila (6pm–2am ET), which covers Caliber's evening deploy windows and the tail of the US business day
Reachable for production incidents during US business hours on a rotation
Paid 2-week trial --> then a 90-day initial term
Rate: $35–$55/hr
About Caliber
Caliber is an inbound lead-generation and call-routing platform serving insurance, energy, home internet and education verticals. It ingests leads from paid media and partner sources, runs a WebRTC softphone and dialer engine for live agents, routes calls and data posts to buyers under real-time eligibility rules, and closes the loop with revenue attribution back to the ad platforms. The stack is TypeScript end to end: Next.js 14 (App Router) on Vercel, Supabase Postgres with RLS, a Node dialer engine and background worker on Fly.io, Telnyx for voice and SMS, and Grafana Cloud for observability. Caliber routes live phone calls, SMS and lead deliveries for buyers who pay per call. By the time you join, a staging environment, gated deploys and automated migrations will exist; they will be new, thin, and owned by one very busy engineer. Your job is to take that operational layer over, harden it, and become the person who is paged first.
What you will do
Own the environment ladder day to day: local, preview, staging, production across Supabase, Fly.io and Vercel. Keep staging trustworthy so every change is proven there first.
Own CI/CD: GitHub Actions for typecheck, lint, tests, migration dry-runs and gated deploys with rollback. Make the engine deploy a button, not a runbook, and run the evening deploy window.
Own Supabase operations: migration application and drift detection, RLS review, backups and quarterly restore drills, partition maintenance on event tables, connection pooling, disk and RAM headroom, storage buckets.
Own Fly.io and Vercel: machine sizing, regions, health checks, rolling deploys for the engine and worker, secrets, log drains.
Build and tune observability: Grafana dashboards and alert rules on engine metrics, worker queue depth, delivery failure rates, Telnyx webhook lag and DB health; route alerts by severity with runbooks you keep current.
Be first responder on pages. Coordinate with the Founding Engineer, write blameless post-mortems, and close the loop with fixes.
Maintain least-privilege access across Supabase, Vercel, Fly, Telnyx and GitHub; rotate secrets; run the on/off-boarding checklist for every contractor.
Keep a monthly cost view of Supabase, Fly, Vercel, Telnyx and AI API spend and flag drift.
Must have
5+ years in DevOps, SRE or platform engineering for production SaaS with real-time or transactional traffic, where you were on the on-call rotation.
Hands-on Postgres operations: migrations, backups and PITR, performance triage, partitioning. Supabase preferred.
GitHub Actions (or equivalent) pipelines you built and maintained, including deploy gating and rollback.
Docker and at least one PaaS at depth: Fly.io preferred; Render, Railway, Heroku or ECS acceptable.
Prometheus/Grafana or Datadog: metrics, dashboards, alert rules, SLOs.
Enough TypeScript/Node to read what you deploy, write a health check, and fix a broken build.
Calm, written-first incident communication in clear English.
Nice to have
Telecom infrastructure: Telnyx or Twilio webhooks, SIP trunks, WebRTC TURN/relay, media-quality telemetry.
Vercel at depth (edge middleware limits, preview environments, env management).
Terraform or Pulumi.
BPO or contact-center platform operations experience.
How we'll evaluate
Written screen (30 min): describe a production incident you handled end to end — detection, what you did in the first ten minutes, what changed afterward.
45-minute conversation with the Founding Engineer: walk through the current pipeline and say what you would harden first and what you would refuse to touch in week one.
Paid 2-week trial: add migration drift detection and a restore drill to the existing pipeline on staging, with a runbook, without production access.
How to apply
Send a short note on the system you have owned that most resembles Caliber, a link to code or a write-up you are proud of, and your rate and availability. We reply to everyone who completes the written screen.

Work arrangement
Yes