Design, implement and evaluate efficient deep neural network architectures and algorithms for d-Matrix's AI compute engine.
Engage and collaborate with internal and external ML researchers to meet R&D goals.
Engage and collaborate with Software team to meet stack development milestones.
Conduct research to guide hardware design.
Develop and maintain tools for high-level simulation and research.
Port customer workloads, optimize them for deployment, generate reference implementations and evaluate performance.
Report and present progress timely and effectively.
Contribute to publications of papers and intellectual property.

Requirements

Pursuing Masters/PhD degree in Computer Science, Electrical and Computer Engineering, or a related scientific discipline.
High proficiency with PyTorch is a must
High proficiency in algorithm analysis, data structure, and Python programming is a must.
Current knowledge in machine learning and modern deep learning.
Hands-on experience with modern neural network architectures such as MoEs and Diffusion models.
Knowledge and experience with efficient deep learning is preferred: quantization, sparsity, distillation.
Strong publication records in top machine learning conferences or journals is preferred.
Proficiency with C/C++ programming is preferred.
Proficiency with GPU CUDA programming is preferred.
Experience with AutoML and meta learning is preferred.
Experience with numerical analysis preferred.
Experience with specialized HW accelerator systems for deep neural network is preferred.

Benefits

Medical/Dental/Vision/401k
Incentive opportunities based on individual and company performance
Diverse package of benefits centered around the wellbeing of employees and their loved ones

Applicant Tracking System Keywords

Tip: use these terms in your resume and cover letter to boost ATS matches.

Hard Skills & Tools

PyTorchalgorithm analysisdata structurePython programmingdeep learningneural network architecturesquantizationsparsitydistillationC/C++ programming

Soft Skills

collaborationcommunicationpresentationresearchtimeliness