AI Inference Framework Engineer

quadric•Burlingame, CA

1d•$160,000 - $195,000

About The Position

Quadric is hiring an AI Inference Framework Engineer for its AI Application team. The role will make Quadric’s General Purpose Neural Processing Unit (GPNPU) a first-class backend in frameworks developers already use, including llama.cpp, PyTorch, vLLM, and Optimum, so customers can move applications from GPUs to Quadric with minimal changes. The engineer will own backend integration from memory management and compute dispatch through serving features, working closely with kernel, runtime, and compiler teams. The work includes integrating the GPNPU backend into AI compute frameworks; implementing framework interfaces, memory allocation, compute dispatch, registration, operators, and kernels; and building model compute pipelines for prefill, decode, batching, multi-sequence, and multimodal workloads. The engineer will support serving features such as KV-cache management, paged attention, continuous batching, and speculative decoding, and optimize framework performance and accuracy. This includes reducing host-device dispatch overhead, implementing graph capture and replay, integrating efficient kernels, supporting quantization formats, and validating accuracy against reference backends. Within the first 6–12 months, success means bringing up a new LLM/VLM family end to end on the GPNPU backend with reference-matching accuracy, delivering measurable tokens-per-second or latency gains on a frontier model, and shipping a serving feature such as paged KV cache in a production release.

Requirements

At least 3 years of experience building or extending ML frameworks or inference runtimes.
BS/MS in Computer Science, Computer Engineering, or a related field.
Strong C/C++ and Python skills.
Solid understanding of framework internals, including tensor memory management, graph scheduling, and operator dispatch.
Solid understanding of AI compute performance analysis and optimization.
Fluency with AI coding agents such as Claude Code or Cursor, and interest in building agent-driven workflows for backend and model bring-up.

Nice To Haves

Experience writing an accelerator backend (CUDA, Metal, Vulkan, Hexagon, or another NPU) for an open-source framework.
Contributions to open-source frameworks such as PyTorch, llama.cpp, or vLLM.
Understanding of low-level AI compute pipelines, CPU/NPU, and DMA.
Understanding of graph compilers and kernel fusion.

Responsibilities

Integrate the GPNPU backend into AI compute frameworks such as llama.cpp, PyTorch, vLLM, and Optimum.
Implement framework backend interfaces, memory allocation, compute dispatch, registration, and operators or kernels.
Implement model compute pipelines for prefill, decode, batching, multi-sequence, and multimodal workloads.
Support serving features including KV-cache management, paged attention, continuous batching, and speculative decoding.
Optimize framework performance and accuracy for the GPNPU backend.
Reduce host-device dispatch overhead, implement graph capture and replay, integrate efficient kernels, and minimize latency.
Support quantization formats such as GGUF (Q4/Q8), GPTQ/AWQ, and int8/fp16.
Validate accuracy against reference backends using measures such as perplexity and task benchmarks, and drive improvements.
Bring up a new LLM/VLM family end to end on the GPNPU backend with accuracy matching the reference.
Deliver measurable tokens-per-second or latency gains on a frontier model and ship a serving feature in a production release.

Benefits

Medical, dental, and vision insurance from day one, with premiums covered at 99% for employees.
Company-paid life insurance.
Voluntary supplemental life insurance.
Short-term disability and long-term disability insurance.
Commuter support, including parking or Caltrain reimbursement.
Flexible spending account (FSA) and health savings account (HSA).
Equity with the business.
Paid parental leave.
401(k) retirement plan.
Flexible PTO.
Winter holiday shutdown.
Catered lunch each day in the office.
Collaborative, low-ego culture with significant ownership and impact.
Work culture focused on innovative disruption.