LiveKit · Backend reliability & self-hosting

Fix the backend around your LiveKit calls.

I investigate call-state, webhook, token and billing bugs in existing LiveKit products. Fixed-scope work, written findings, and changes your team can maintain.

Describe your setup See the work

Node.js & Go · Written scope before work starts

Already self-hosting or weighing a move from Cloud? Start with a deployment audit or a hosting cost assessment.

For a product that already has calls

  • Your call state disagrees with room activity: early joins, disconnects or retries leave users in the wrong state.
  • A call affects money or access, and the billing or token flow needs a reproducible test rather than another patch based on guesswork.
  • Your team or agency needs an engineer for a defined backend problem, with code, tests and a written handover.

These backend investigations work with Cloud or self-hosted deployments. A hosting change is a separate decision, not a prerequisite.

What I investigate

One call flow, traced across the boundaries that matter to your product.

Call lifecycle: creation, joining, disconnect, reconnect and finalization.
Webhook handling: signature checks, retries, duplicate processing and recovery when events are missing.
Access: whether token issuance respects the caller, the room and the current business state.
Billing: whether chargeable time follows the agreed call rules, including pauses and termination.

LiveKit retries webhooks but does not guarantee delivery. A backend needs a plan for those gaps. LiveKit delivery documentation.

Start with a bounded piece of work

Call-flow audit

from $3k · fixed scope

  • One agreed call flow: application code, relevant events and available logs
  • A reproduction where the supplied evidence permits it; otherwise a documented evidence gap and next test
  • Written findings separating confirmed faults, hypotheses and recommended changes

What you keep: the evidence timeline, prioritized findings and a regression-test plan. Implementation is quoted separately.

Backend fixes & integration

quoted after scoping

  • A defined Node.js or Go change, based on an audit or an existing reproducible issue
  • Acceptance criteria and regression tests agreed before implementation
  • Code review, rollout and rollback notes for your team

What you keep: code changes, tests and a written handover in your repositories. Production rollout requires your approval.

The quote sets the engineering time, required access, deliverables and schedule. No open-ended investigation, automatic hosting migration or bundled on-call commitment.

Describe your setup

Self-hosted LiveKit: review or migration

Keep an existing deployment healthy, or decide whether moving from Cloud makes sense for your workload and operating costs.

Self-hosted deployment audit

from $3k · fixed scope

  • Review the agreed configuration, network paths and observability
  • Test a defined workload or failure scenario, with explicit limits
  • Get written findings, unresolved risks and a prioritized fix and retest plan

Implementation is separate. See the self-hosted audit scope.

Cloud → self-hosted assessment

assessment quoted after scoping

  • Compare your Cloud bill with compute, bandwidth, storage and operating effort
  • Check dependencies, missing features and who will run the deployment
  • Get a stay-or-move recommendation before committing to a migration

If justified, migration implementation starts from $5k under a separate scope. See the cost assessment and migration process.

Self-hosting is not a savings guarantee. The comparison includes migration effort, redundancy, upgrades and incident response. Start with the self-hosting validation checklist if you want to prepare the evidence first.

Work behind the offer

Application-level work on a creator platform using LiveKit Cloud and per-minute video billing. Anonymized summaries, not self-hosted infrastructure benchmarks.

$ cat call-lifecycle-cases.log
01 · billing problem: billing started before both participants joined
change: tie billing start to participant presence, not call creation 02 · disconnect problem: charging continued after a participant disconnected
change: add presence-driven pause/resume handling to billing 03 · early join problem: issuing a join token changed call state and prematurely finalized a scheduled call
change: separate token issuance from call-start transitions 04 · access problem: the token endpoint still issued tokens for cancelled calls
change: reject token requests for finalized call states

I’m an independent backend engineer working with Node.js and Go. My LiveKit production work is at the application layer: rooms, webhooks, join tokens and billing synchronization. My self-hosted work is a separate lab, not a claim of operating client SFU fleets.

Lab work: livekit-prometheus-alerts — alert rules, a metrics inventory and a Grafana dashboard. livekit-idle-room-reaper — a dry-run-first cleanup tool for selected room and recording cases. Check native timeouts, version compatibility and the documented limits before using either.

FAQ

Do I need to leave LiveKit Cloud?

No. I work on the application backend around your existing LiveKit setup. LiveKit Cloud now offers EU data-residency controls; EU residency alone is not a reason to assume self-hosting is required. Check the current documentation and DPA for your data flows and requirements. Moving infrastructure is a separate decision, not a prerequisite for a call-flow review.

What is your production experience?

My production experience is the application layer of LiveKit Cloud: rooms, webhooks, join tokens and per-minute video billing, with Node.js and Go backend work. My self-hosted monitoring and egress work is lab work, published separately. I do not present that as operating client SFU fleets or production SIP systems.

What is outside the scope?

This is not a 24/7 on-call service, a compliance certification or a promise to reach a particular speech-recognition score. Carrier/SIP troubleshooting, model tuning and infrastructure migrations are not included in the call-flow audit. If the evidence points outside my scope, I document that boundary rather than promise a fix.

What do you need to start?

Email your stack, the failure or hosting decision you need help with, its business impact and your deadline. For a cost assessment, include rough usage and spend without account or customer details. No credentials, recordings or personal data in the first message. If the task fits, we agree the scope, deliverables, price and start date in writing.

Who does the work, and what stays with us?

You work directly with an independent backend engineer. Each engagement includes written findings and a handover for the agreed scope. For implementation work, your team receives the code changes, tests and operating notes. Your accounts and repositories stay under your control.

Describe your setup