Infrastructure Engineer (Founding Team) in San Francisco

• End-to-End Stack Ownership: Architect, scale, and maintain multi-cloud GPU infrastructure across AWS, GCP, Base10, and AWS Marketplace.
• Low-Latency Inference: Design and deploy ultra-fast, high-reliability GPU serving systems handling live customer traffic.
• Enterprise & On-Prem: Deploy, manage, and optimize custom infrastructure setups for tier-1 enterprise clients.
• Efficiency & Reliability: Continuously optimize GPU utilization, system scaling, and cloud expenditure.

• Tier-1 Technical Pedigree: Strong Computer Science/Engineering foundation from a top-tier institute with deep systems-level depth.
• Startup & Zero-to-One Track Record: 2+ years of hands-on infrastructure experience with proven architectural ownership at a high-growth, venture-backed startup or high-value tech company. Must have built scalable infrastructure systems from scratch (not just managed narrow slices like training pipelines).
• High Ownership & Velocity: Ability to thrive in an intense, high-speed, in-person startup environment (SF Hacker House setup).
• Location: Based in or willing to relocate to San Francisco (Visa sponsorship available: H-1B, O-1, OPT).
• This is a full-time, permanent core team role. Applications for contract, C2C, consulting, agency, or short-term engagements will be immediately rejected.

We are an elite, high-growth AI research and infrastructure company building next- LLM interpretability and context optimization systems. Our custom ML models analyze and compress token contexts before they hit underlying models-slashing inference costs by ~50%, dropping latency, and boosting accuracy.

Just seven months post-launch with over 1,000 paying customers, we are backed by $11.7M from First Round Capital, Y Combinator, and the founders of Hugging Face, Slack, and Dropbox.

Company DescriptionWe are an elite, high-growth AI research and infrastructure company building next- LLM interpretability and context optimization systems. Our custom ML models analyze and compress token contexts before they hit underlying models-slashing inference costs by ~50%, dropping latency, and boosting accuracy.\n\nJust seven months post-launch with over 1,000 paying customers, we are backed by $11.7M from First Round Capital, Y Combinator, and the founders of Hugging Face, Slack, and Dropbox.