Member of Technical Staff, Infrastructure

Vapi · San Francisco, California · Posted 2026-05-21

Vapi (/ˈVɑːpi/):

Powering 1 billion calls for companies like Amazon Ring, Intuit, ServiceTitan, and New York Life

Backed by Peak XV, Bessemer, Kleiner Perkins, M12, Y Combinator, and more with $72M raised

Why We’re Hiring This Role: -------------------------------

Vapi runs live phone calls — when something breaks, callers hear it. We’re building cell-based, multi-region infrastructure to drive 99.99% call completion, and this hire owns the foundation: multi-cluster Kubernetes on EKS, a stateful data plane (Postgres, Redis, Kafka, Temporal, ClickHouse), Envoy/Cilium networking, and multi-region Kafka on MSK across EU and ANZ.

You’ll write Go for control-plane services like cluster-manager, traffic-control-plane, and environment-manager, and you’ll set the bar for how Vapi runs stateful workloads at scale.

What You’ll Do: -------------------

30 Day: Ramp on the cell-based architecture, the regional EKS clusters (backend / networking / persistence / monitoring / models / kafka), and the Pulumi stacks. Shadow oncall, walk recent incidents (Envoy response flags, conntrack drops, cross-zone LB target resets), and own a first scoped infra change end-to-end.

60 Day: Take ownership of one core domain — e.g., multi-region MSK (regional topic naming, Pulumi drift, compliance constraints), the Postgres/Neon consolidation path, or programmatic cluster creation via Cluster API. Ship a control-plane improvement in Go and drive a measurable reliability or capacity win.

90 Day: Lead a roadmap pillar of the cell-based build-out: a new region, a stateful workload migration, or unblocking the SIP gateway SPOF. Operate as the infra owner other teams pull in for design reviews, and set the standards (runbooks, failure-domain modeling, capacity targets) the next infra hires inherit.

Who You Are: ----------------

You’ve run multi-cluster Kubernetes on EKS in production — backend, networking, persistence, monitoring, models, and kafka clusters per region — and you’ve used Cluster API or similar for programmatic cluster creation.

You’ve operated a stateful data plane (Postgres, Redis, Kafka, Temporal, etcd, ClickHouse) at scale — you’ve sharded it, migrated data between instances, and lived with the consequences.

You’re fluent in Envoy and Cilium/eBPF. You’re comfortable debugging Envoy response flags, conntrack drops, and cross-zone LB behavior. VPC/NAT/Cloudflare alone isn’t enough.

You’ve run multi-region Kafka on MSK in production — not just Kafka. You’ve dealt with regional topic naming, MSK Pulumi drift, and compliance constraints.

You write Go for control-plane services. Vapi’s cluster-manager, traffic-control-plane, and environment-manager are all Go, and you’re comfortable owning code in that stack.

Bonus: SIP / RTP / telephony background. The Nov 7 SIP gateway SPOF is still unsolved, and a telephony-savvy infra hire unblocks that roadmap item.

Bonus: cell-based / shard architecture experience — Shopify pods, AWS cell-based reference arch, Slack shards, or equivalent. Microservices experience alone isn’t the same.

You likely come from one of: a company that ran cell-based in prod (Shopify, AWS service teams, Slack); a distributed systems shop (Cockroach, MongoDB, Confluent, Temporal, Redpanda, ClickHouse Inc.); a voice/video/CPaaS company (Twilio, Plivo, Bandwidth, Vonage, LiveKit, Daily.co, Dialpad); an Envoy/service-mesh org (Lyft, Stripe, Airbnb, Pinterest, Isovalent/Cilium); or a streaming-infra team (Confluent, Uber, LinkedIn, Datadog) that ran MSK/Kafka multi-region.

Why Vapi: -------------

What We Offer: ------------------

About Vapi

Voice AI for developers.