Etched.ai is an AI infrastructure company building rack-scale and frontier inference systems.
About Etched Etched is building hardware for frontier intelligence. We co-design chips, racks, software, and manufacturing to deliver best-in-class throughput and latency across both prefill and decode workloads. Our first products are heavily focused on inference. Backed by hundreds of millions from top-tier investors and staffed by leading engineers, Etched is redefining the infrastructure layer for the fastest growing industry in history. Job Summary Every model release presents a new opportunity to push the frontier on kernel engineering. Future performance breakthroughs will come from AI systems that can understand model architectures and hardware, run thousands of experiments, learn from compiler and profiler feedback, and discover the most performant implementations faster than the best engineers. You will build that system. Your mandate is to build AI systems that autonomously turn newly released model architectures into correct, production-ready implementations optimized for Etched hardware. These systems should explore broader design spaces, learn from every experiment, and reach peak performance faster than any traditional kernel-development workflows. Etched offers a uniquely tight research loop: proprietary hardware, compiler, runtime, kernels, production workloads, and dedicated in-office compute under one roof. You will teach models using proprietary performance signals, iterate on their proposals, and make every experiment improve both the performance optimization system and the hardware it runs on. Key Responsibilities - Own the system that turns new model architectures into verified, production-ready kernels and model mappings. - Build agents that understand Etched hardware, design experiments, generate implementations, compile and profile them, diagnose bottlenecks, and iterate with our teams, to the limits of model autonomy. - Design evals covering correctness, numerical stability, latency and efficiency. - Turn profiler traces, simulation, hardware counters, and expert judgment into structured signals models can learn from. - Curate proprietary datasets and optimization memory from complete trajectories, expert demonstrations, counterexamples, and production outcomes. - Build fast, reproducible experiment infrastructure and observability so experiments remain interpretable, trustworthy, and high-throughput. - Ship model-generated improvements to production and quantify their impact on end-to-end system performance. - Partner deeply with other architecture teams to shape new abstractions and Etched’s hardware-software roadmap. - Continuously evaluate new model releases and deploy the best for each stage of the optimization loop. You may be a good fit if you have - A track record of solving hard problems across stacks and domains — you enjoy being dropped into unfamiliar territory and figuring it out - Comfort with both Python and low-level code: you can read it, modify it, debug it, and direct AI to write it well. We do not care whether you write code from scratch — we care whether you ship things that work. - Kernel experience: you've written or tuned kernels and can explain the mechanisms and performance impact of optimizations you’ve shipped - Fluency using AI to learn and ramp on new problems — agentic coding tools, deep research, and frontier models are how you work, not an add-on - Moving fluidly between research exploration, agentic experimentation, low-level debugging, and production execution. Strong candidates may also have experience with - First principles thinking on accelerator performance: memory hierarchy, data movement, parallelism, synchronization, and low-precision computation. - Hands-on experience building and shipping LLM-based agents or AI tooling that real users depend on in production environments (beyond calling an API — context engineering, tool integration, orchestration, failure analysis) - An eval-driven mindset: you measure whether AI systems work before scaling them - Fine-tuning or post-training, RAG over proprietary data, and/or multi-agent orchestration - High agency and comfort with ambiguity — you find the real problem to solve Benefits - Medical, dental, and vision packages with generous premium coverage - $500 per month credit for waiving medical benefits - Housing subsidy of $2k per month for those living within walking distance of the office - Relocation support for those moving to San Jose (Santana Row) - Various wellness benefits covering fitness, mental health, and more - Daily lunch and dinner in our office - Unlimited compute budget subject to ROI justification Base Compensation Range - $150,000 – $225,000 How we’re different Etched believes in the Bitter Lesson http://www.incompleteideas.net/IncIdeas/BitterLesson.html. We are the first inference-focused frontier AI system, betting early on transformer and transformer-like architectures and on increasing model sizes. Our addressable market is the entirety of inference, unlike many of our competitors. We are a fully in-person team in San Jose (Santana Row), and greatly value engineering skills. We do not have boundaries between engineering and research, and we expect all of our technical staff to contribute to both and work across disciplines as needed.