Staff Deep Learning Compiler Engineer

Job Title

Staff Deep Learning Compiler Engineer

Role Summary

Design and optimize the deep learning compiler stack (MLIR, TVM, LLVM) to map modern neural networks efficiently to Quadric's GPNPU hardware. Work with software and hardware teams to enable high-performance, low-latency, memory-efficient inference on edge devices.

Experience Level

Senior level - requires significant prior experience; the role expects 8+ years of relevant experience.

Responsibilities

Deliver compiler infrastructure, model lowering, and performance optimizations to maximize hardware utilization and efficiency.

• Design, implement, and maintain compiler optimization passes for Quadric’s architecture using MLIR, Apache TVM, LLVM, or similar frameworks.
• Develop lowering paths from PyTorch/TensorFlow/ONNX into optimized kernel code and codegen.
• Implement graph-level optimizations: operator fusion, layout transforms, quantization strategies, and memory allocation.
• Profile and benchmark models (Transformers, CNNs, VLMs) to identify and resolve compiler bottlenecks.
• Collaborate with hardware and micro-architecture teams to define ISA extensions and hardware acceleration features.
• Build software simulators, functional models, and test benches; support FPGA/ASIC bring-up and validation.

Requirements

Must-have skills and experience to perform the role; preferred items listed separately.

• 8+ years hands-on experience developing deep learning compilers or compiler infrastructures (MLIR, Apache TVM, LLVM, XLA, TensorRT, Glow, or similar).
• Strong C++ (C++14/17/20) and Python proficiency; solid fundamentals in data structures, algorithms, and OOP design.
• Familiarity with modern ML frameworks (PyTorch, TensorFlow, ONNX) and operator representations.
• Deep understanding of computer systems: memory hierarchies, parallel processing, and instruction execution on CPU/GPU/NPU.

Preferred / nice-to-have:

• Low-level kernel optimization experience, SIMD/vector programming, and memory allocation strategy development.
• Experience with model quantization (INT8, FP8, mixed-precision), post-training and QAT workflows.
• Prior work on software stacks for custom AI accelerators, DSPs, or embedded architectures.
• Experience with FPGA bring-up, hardware emulation, or cycle-accurate simulator development.

Education Requirements

BS, MS, or PhD in Computer Science, Computer Engineering, Electrical Engineering, or a related field; equivalent practical experience likely considered.

About the Company

Company: Quadric

Headquarters: Burlingame, California, United States

Quadric is building the world’s first supercomputer designed for the real-time needs of edge devices. Founded in 2016, the company empowers developers across industries with innovative general-purpose neural processing unit (GPNPU) architecture for neural network workloads. Co-founded by technologists from MIT and Carnegie Mellon, Quadric aims to enable groundbreaking technology development.