Member of Technical Staff — Reliability-CI Infrastructure

RadixArk; Palo Alto (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

RadixArk is seeking a Member of Technical Staff - CI / Infrastructure to own and architect the CI system that powers SGLang. You will replace brittle thresholds with regression-based detection, harden runners against fork PRs, and cut cycle times so contributors get feedback in minutes.

This role emphasizes designing CI infrastructure as a system you treat with the same care as serving infrastructure, collaborating with core maintainers and hardware partners to keep the gating process fast,

Qualifications

• 3+ years operating CI/CD at scale.
• Deep Linux, Docker, GPU computing knowledge.
• Self-hosted runner management experience.
• Strong Bash and Python.
• Security mindset - CI supply chain risks, fork PR attack vectors, runner hardening.
• NVIDIA GPU drivers, CUDA, NCCL, InfiniBand/RDMA experience in CI contexts.
• Familiarity with ML inference workloads (model loading, KV cache, quantization).

Responsibilities

• Own CI reliability end-to-end - triage failures, distinguish real regressions from flaky tests and infra issues, keep main green.
• Build regression-based CI - replace hardcoded static thresholds with automated baseline comparison (metrics pipeline, durable storage, detection logic).
• Harden runner infrastructure - ephemer...
• Cut CI time - right-size eval suites, deduplicate server startups, separate PR smoke tests from nightly full runs.
• Improve developer experience - faster feedback, clearer failure messages, workflow orchestration.

Skills

CI/CD at scale

Linux

Bash

Python

Security mindset

Self-hosted runner management

GPU computing basics

Tools

GitHub Actions

Buildkite

Jenkins

GitLab CI

Docker

CUDA

NCCL

InfiniBand/RDMA

Job description

About the Role

RadixArk is hiring a Member of Technical Staff - CI / Infrastructure to own the infrastructure that keeps SGLang moving. Our CI system runs 300+ GPU tests across NVIDIA, AMD, Intel, and Ascend hardware pools, gating every commit to one of the fastest-growing open-source LLM inference engines. When CI is green and fast, 100+ contributors ship with confidence. When it isn't, the entire project stalls. That bottleneck is your problem to solve.

You won't just maintain pipelines - you'll architect them. You'll replace brittle static thresholds with regression-based detection, harden runners against supply-chain attacks from fork PRs, and cut cycle times so contributors get feedback in minutes, not hours. You'll work directly with core maintainers, hardware partners, and the open-source community to keep the system that gates every merge request trustworthy, fast, and secure.

This is not a role for someone who wants to write CI YAML and walk away. It's for an engineer who treats CI infrastructure the way we treat serving infrastructure - as a system worth designing well.

What You'll Do

• Own CI reliability end-to-end - triage failures, distinguish real regressions from flaky tests and infra issues, keep main green

• Build regression-based CI - replace hardcoded static thresholds with automated baseline comparison (metrics pipeline, durable storage, detection logic)

• Harden runner infrastructure - ephemer...

• Cut CI time - right-size eval suites, deduplicate server startups, separate PR smoke tests from nightly full runs

• Improve developer experience - faster feedback, clearer failure messages, workflow orchestration

Requirements

• 3+ years operating CI/CD at scale (GitHub Actions, Buildkite, Jenkins, GitLab CI, or similar)

• Deep Linux, Docker, GPU computing knowledge

• Self-hosted runner management experience

• Strong Bash and Python

• Security mindset - CI supply chain risks, fork PR attack vectors, runner hardening

• NVIDIA GPU drivers, CUDA, NCCL, InfiniBand/RDMA experience in CI contexts

• Familiarity with ML inference workloads (model loading, KV cache, quantization)

Nice to Have

• Large open-source project CI experience (100+ contributors)