Distributed DL Algorithms Developer
Define high level parallelism semantics and APIs for building models in nervana graph framework and frontends (neon, TensorFlow, …)
Implement sync/async SGD, parameter servers, and other state of the art distributed DL algorithms
Profile distributed DL models to identify performance bottlenecks and propose solutions across individual component teams
Transform computational graph representations of neural network models to express parallelism and communication dependencies
Build and maintain runtime framework for distributing and executing model subgraphs across hosts or processors
Integrate low-level communication libraries (MPI, cuda/nccl, etc.) with backend implementations of nervana graph
Interacting with deep learning researchers to support different frontend frameworks
Interacting with device driver developers and hardware architects to influence low level features
M.S. or Ph.D. in CS, ECE or related fields and minimum 4 years of software development experience
Strong statistics, calculus and applied math skills
Background in applied deep learning, using existing frameworks and state of the art models
Proficient in Python and C
Confident in SW engineering, design and testing practices
Familiar with collective communication patterns and usage in distributed algorithms
Experience with multiprocessing
Inside this Business Group
Intel Nervana, leveraging Intel's world leading position in silicon innovation and proven history in creating the compute standards that power our world, is transforming Artificial Intelligence (AI). Harnessing silicon designed specifically for AI, end-to-end solutions that broadly span from the data center to the edge, and tools that enable customers to quickly deploy and scale up, Intel Nervana is inside AI and leading the next evolution of compute.
Posting Statement. Intel prohibits discrimination based on race, color, religion, gender, national origin, age, disability, veteran status, marital status, pregnancy, gender expression or identity, sexual orientation or any other legally protected status.
A little about us:
Our vision is simple and direct. If it computes, it does it best with Intel. We embrace all aspects of computing.