Machine Learning Scientist: Co-design

The AI age is upon us and high performance computing is the underlying platform powering everything from Large Language Models to Image synthesis from text. However, with the demise of Moore’s laws and Dennard scaling we are at an inflection point. At Lightmatter, we are leading the transition of computing from traditional electronic transistors to photonic technologies which can operate at mind blowing efficiency and throughput.

In this role, you will guide the development of groundbreaking holistic solutions that combine the strengths of photonics, ASICs, software and ML algorithms. It is an opportunity to architect distributed computing approaches such as tensor parallel computing with the connectivity and the memory hierarchy of high performance ASICs. Working in a small team of highly talented engineers, you will be able to move fast and develop well architected solutions.

If you are passionate about advanced AI technology and would like to develop scalable algorithms, hardware and ML techniques, join us!

Responsibilities

• Develop parallel algorithms for balancing compute and communication (within an accelerator or between accelerators) to maximize throughput and minimize latency.
• Deliver hardware and software co-design targeted at low-latency inference.
• Influence the development of machine learning hardware by simulating low-latency and high throughput inference for different models.
• Publish and present new research at premier ML/CS conferences.

Requirements

• MS in Computer Science or related fields; PhD strongly preferred
• 10+ years of industry experience
• Expert understanding of deep learning, parallel computing, compilers and/or hardware architecture.
• Experience in working with large ML/HPC workloads with distributed computing systems built with accelerators such as GPUs or TPUs.
• Experience with developing and modifying machine learning models for scalability.
• Experience or understanding of low precision training and inference.

Technical expertise

• Strong technical understanding of advanced techniques used in parallel computing, deep learning and HPC.
• Ability to model complex workloads on different architecture proposals.
• Understanding of parallel computing architectures.
• Experience with scalable frameworks such as MPI, PyTorch distributed, CUDA, and NCCL.
• Highly proficient in deep learning programming languages and frameworks, e.g. Python, C++, CUDA, Tensorflow, PyTorch, JAX.
• Experience with practical problem solving with innovative algorithmic solutions.

Preferred qualifications

• Experience contributing first hand to important software or ML algorithms deployed in the industry.
• Experience working with compiler optimizations is a plus.
• Strong publication records in the field of machine learning, parallel computing, and/or computer architecture.
• You have demonstrated the ability to perform independent research.
• Prior experience with quantization and compression methods within deep learning.
• You are enthusiastic about new technologies, algorithms, and mathematics.

Benefits

• Health Care Plan (Medical, Dental & Vision)
• Short Term Disability Insurance
• Retirement Plan 401k
• Life Insurance
• Paid Time Off
• Sick day
• Family Leave
• Stock Option Plan

Base Compensation Range: $220,000 to $230,000. In accordance with the Colorado, California and New York law, the range provided is Lightmatter's reasonable estimate of the compensation for this role. Actual pay will be based on several factors including work experience, location and education.

Lightmatter recruits, employs, trains, compensates and promotes regardless of race, religion, color, national origin, sex, disability, age, veteran status, and other protected status as required by applicable law.