СШАDeep Learning Compiler Intern
1 нед. назад
СШАTraineeOnsite
c++compilerneural networkdebugging
Internship role working on neural network compiler passes and code generation for 's GPNPU hardware.
Обязанности
- As a Deep Learning Compiler Intern, you will work closely with our senior compiler engineers on CGC, 's neural network compiler that lowers to code targeting the Chimera GPNPU. You will dig into real compiler passes — layout selection, memory allocation, operator splitting, code generation — and see your changes flow end-to-end into the C++ that runs on silicon. This is a hands-on role where you will gain experience designing IR transformations, debugging generated code, and improving how efficiently neural networks map to hardware.
- Note: Our preference is for this internship to be based out of our Burlingame, California office. Candidates should be based in the Bay Area or able to relocate for the internship period and available to work on site.
- Develop & Implement
- Help build and extend compiler passes that lower neural network IR to GPNPU-targeted code.
- Analyze & Debug
- Diagnose compilation issues by tracing problems from generated C++ back through the pipeline. Use IR dumps and static analyses to investigate compilation failures and performance regressions.
- Code Optimization
- Work alongside senior engineers to improve compiler decisions to reduce data movement and increase core utilization.
- Collaborate
- Partner with the kernel, hardware, and data science teams to align compiler features with real model requirements and hardware constraints.
- Toolchain Contribution
- Contribute to test infrastructure, debugging utilities, and developer ergonomics across the CGC pipeline and runtime.
- Within your internship, you'll:
- Build a working understanding of how neural networks are compiled and mapped to GPNPU hardware.
- Contribute compiler passes, debugging tools, or performance improvements that meaningfully support the CGC pipeline.
- Develop hands-on fluency with IR transformations and hardware-aware code generation.
Требования
- Required
- Currently pursuing a Bachelor's, Master's, or PhD in Computer Science, Electrical Engineering, or a related field.
- Strong proficiency in Python and C++.
- Foundational understanding of compiler concepts: intermediate representations, dataflow analysis, and transformation passes.
- Comfort reading and reasoning about large, unfamiliar codebases.
- Demonstrated capability in problem-solving, debugging, and clear technical communication.
- Preferred
- Coursework or project experience with compilers, program analysis, or domain-specific languages.
- Exposure to ML compiler frameworks such as TVM, MLIR, XLA, Glow, or IREE.
- Familiarity with neural network quantization, fixed-point arithmetic, or numerical analysis.
- Experience with hardware-aware code generation for accelerators (GPU, DSP, NPU).
- Some exposure to assembly or low-level code generation.
- Previous internship experience in compilers, ML systems, or performance engineering.
Условия
- The hourly salary range for this temporary position is $50-57/hr . This range reflects the full span of levels and geographies at which hires for this role. The actual hourly salary offered will depend on a number of factors, including the specific level of the role, years and depth of relevant experience, technical skills and competencies.
- interns receive hands-on experience working alongside industry experts in AI and semiconductor technology, with access to mentorship and meaningful project ownership from day one along with following
- Catered lunch each day in our office
- Downtown Burlingame office location, close to shops, cafes, and local amenities
- A work culture focused on innovative disruption
- If this role resonates with you, we encourage you to apply even if your experience does not perfectly match every qualification. We value potential, curiosity, and a willingness to learn just as much as direct experience.
Другое
- is redefining edge AI with the industry's first General Purpose Neural Processing Unit (GPNPU), enabling developers to run both neural network inference and conventional C++ code on a single programmable architecture. Our technology powers intelligent edge devices across automotive, industrial, robotics, and embedded systems.
- Founded in 2016 and based in downtown Burlingame, California, is building the world's first supercomputer designed for the real-time needs of edge devices. aims to empower developers in every industry with superpowers to create tomorrow's technology, today. The company was co-founded by technologists from MIT and Carnegie Mellon, who were previously the technical co-founders of the Bitcoin computing company 21.
- By submitting an application, you acknowledge that will collect and process your personal information as part of the hiring process. Please review our Privacy Policy to understand how we handle your data