Quadric has created an innovative general purpose neural processing unit (GPNPU) architecture. Quadric's co-optimized software and hardware is targeted to run neural network (NN) inference workloads in a wide variety of edge and endpoint devices, ranging from battery operated smart-sensor systems to high-performance automotive or autonomous vehicle systems. Unlike other NPUs or neural network accelerators in the industry today that can only accelerate a portion of a machine learning graph, the Quadric GPNPU executes both NN graph code and conventional C++ DSP and control code.
As a Senior Performance Architect, you will be the critical link between software and hardware, responsible for understanding how code executes on Quadric's architecture and identifying opportunities for optimization. You will analyze workloads from high-level C++ and Python down through generated assembly to pinpoint performance bottlenecks. This is a hands-on role: beyond analysis, you will prototype solutions yourself - whether that means writing optimized code, modifying compiler passes, or building proof-of-concept implementations to validate proposed fixes before handing off to the appropriate team for productization.
This role requires regular work from the Quadric office in Burlingame, CA, a minimum of 2–3 days per week, with some weeks requiring more days onsite based on business needs. Candidates must be able to commute to the office.
Responsibilities- Analyze application performance across the full stack: C++/Python source, compiler output, assembly, and hardware execution
- Identify and localize performance bottlenecks to specific code regions, assembly sequences, or architectural limitations
- Implement proof-of-concept fixes and optimizations to validate proposed solutions before broader rollout
- Develop and maintain profiling infrastructure, benchmarks, and performance regression tests
- Collaborate with compiler engineers to improve code generation and optimization passes
- Work with hardware architects to identify microarchitectural improvements and validate performance models
- Create performance models that predict workload behavior and guide optimization priorities
- Document findings and communicate performance insights to both technical and non-technical stakeholders
- Support customer engagements by analyzing their workloads and recommending optimizations
Requirements
- BS/MS in Computer Science, Computer Engineering, or Electrical Engineering with 5+ years of performance analysis experience
- Strong proficiency in C++ and Python; ability to read, reason about, and write optimized code at the assembly level
- Hands-on mentality: comfortable implementing proof-of-concept solutions, not just identifying problems
- Deep understanding of computer architecture: pipelines, caches, memory hierarchies, SIMD/vector execution
- Experience with profiling tools (perf, VTune, custom trace analysis) and performance debugging methodologies
- Ability to trace performance issues from application behavior down to microarchitectural root causes
- Strong analytical and problem-solving skills with attention to detail
- Excellent communication skills; ability to explain complex performance issues to diverse audiences
- Experience working cross-functionally with compiler, runtime, and hardware teams
- Experience with ML/AI workloads and frameworks (PyTorch, TensorFlow, ONNX)
- Background in compiler development or code generation
- Experience with GPU, DSP, or custom accelerator architectures
- Familiarity with cycle-accurate simulation and performance modeling tools
- Establish systematic performance analysis methodology and tooling for Quadric's software stack
- Identify and drive resolution of top performance bottlenecks in key customer workloads
- Build performance models that accurately predict workload behavior within 10-15% of actual measurements
- Become the go-to expert for performance questions spanning the hardware/software boundary
Benefits
- Competitive salary and meaningful equity
- Medical, dental, and vision coverage starting on day one
- 401(k) retirement plan
- Flexible paid time off (unlimited, non-accrual) to support work-life balance
- When working in-office, enjoy company-provided lunches and a stocked kitchen
- Convenient office location within walking distance of the Caltrain station
- Support for commuting, including monthly parking or Caltrain passes
- Downtown Burlingame office location, close to shops, cafes, and local amenities
- A politics-free, highly collaborative environment where talented people can do their best work and make an immediate impact
- The opportunity to build long-term career relationships in a company that values strong personal connections alongside professional excellence
Top Skills
What We Do
Quadric has built a unified hardware/software architecture optimized for on-device machine learning inference. Only the Quadric GPNPU (general purpose neural processing unit) delivers high ML inference performance while also running C++ code without forcing the developer to artificially partition application code between two or three different kinds of processors. Quadric's GPNPU is a licensable processor IP core that scales from 1 to 64 TOPs and seamlessly intermixes scalar, vector and matrix code.
