NPU Compiler - SW Specialist

Samsung Electronics

Samsung Electronics

Bengaluru, Karnataka, India · Remote
Posted on Sunday, December 12, 2021

Position Summary

1. Explore areas to apply machine learning technology, develop measures on collecting data used for model learning and evaluation and collect data.
2. Develop data pre-processing and feature extraction algorithms and modeling/pattern recognition algorithm by reflecting data characteristics.
3. Perform to optimize data analysis software, develop performance evaluation measures and evaluate.

Role and Responsibilities

Samsung’s NPU (Neural Processing Unit) software team provides exciting opportunity for software engineers to be at the heart of enabling, optimizing and deploying deep learning algorithms on Samsung’s Neural Processing architecture. Exciting new applications in multiple verticals are expected to exploit the power of neural networks and change many aspects of daily lives. This role will require development and optimization of compilers for running the latest deep learning algorithms on specialized hardware i.e. neural processing unit for best performance and power.


  • Experience with C++ 14/17 (advanced features), Python development
  • Hands-on Experience with OOD, TDD development solutions
  • Experience in profiling software and software Optimization techniques
  • Experience in machine learning frameworks and neural network architectures ML compiler workload synthesis [Preferable]
  • Prior working experience in hardware accelerators and hardware-software co-design
  • LLVM or any industrial-strength compiler development experience is a plus.
  • Knowledge of the architecture and understanding of compiler internals [Preferable]
  • Good understanding of System Software, SoC architecture.

Skills and Qualifications

A minimum of 5 to 10 Years

* Please visit Samsung membership to see Privacy Policy, which defaults according to your location. You can change Country/Language at the bottom of the page. If you are European Economic Resident, please click here.