Logo of Huzzle

Home

Matches

Explore

Manage

2025 Intern, AI Algorithm NPU Simulation

image

Samsung Research America (SRA)

1mo ago

  • Internship
    Full-time
    Off-cycle Internship
  • Software Engineering
    IT & Cybersecurity
  • Mountain View
  • Quick Apply

AI generated summary

  • You should have an MS/PhD in a relevant field, strong programming skills in C/C++, Python, and MATLAB, experience with AI libraries, and knowledge of NPU architecture and image quality metrics.
  • You will build and simulate NPU models, analyze AI algorithms, create neural network operators, quantize models, and identify hardware bottlenecks while communicating findings to engineering teams.

Requirements

  • Pursuing or completed MS/PhD degree in Computer Science, Electrical/Computer Engineering, or related discipline
  • Proficient in C/C++, Python and MATLAB programming languages
  • Hands-on experience with Github, PyTorch, TensorFlow, Visual Studio, ONNX libraries
  • In-depth understanding of camera enhancement algorithms, computational photography, CNN, transformer, diffusion, LLM/LVM, AI model quantization
  • NPU hardware micro architecture, on-chip memory, DMA, memory compression, SIMD
  • Extensive experience on implementing prototype code
  • Strong grasp of memory and computation requirements and image quality expectations of state-of-the-art algorithms
  • Experience on identifying image/video quality metrics
  • Effective verbal and written communication skills to work with people across a wide range of disciplines and experience

Responsibilities

  • Work with team members to build NPU simulator which mimics the behavior of NPU hardware within a mobile SoC (System on a Chip) at cycle-accurate or bit-accurate level
  • Work with the algorithm team to cover the use cases including camera, video, diffusion, LLM, LVM applications
  • Analyze AI algorithms and create neural network operators to NPU simulator
  • AI algorithm quantization to reduce model size while maintaining model accuracy
  • Create scheduling mechanism (compiler) to distribute AI model tasks to NPU hardware or simulator
  • Simulate AI algorithms on NPU hardware or simulator and identify the bottlenecks of hardware, such as memory bound or compute bound, and make optimal hardware proposal to mitigate the bottlenecks
  • Communicate the results across related engineering audiences (SW, HW, Architecture, Leadership)

FAQs

What educational qualifications are required for this internship?

Candidates are required to be pursuing or have completed an MS/PhD degree in Computer Science, Electrical/Computer Engineering, or a related discipline.

What programming languages should I be proficient in for this position?

Proficiency in C/C++, Python, and MATLAB programming languages is required for this internship.

Is experience with AI algorithm development necessary?

Yes, hands-on experience with AI algorithms, particularly in camera image and video enhancement, as well as knowledge of NPU software/hardware, is required.

Will I be working alone or as part of a team?

You will be working as part of a collaborative team to build the NPU simulator and analyze AI algorithms.

What types of projects will I be working on?

Interns will work on projects related to building a programmable NPU simulator, analyzing AI algorithms, and implementing neural network operators for various applications including camera and video enhancements.

What tools and libraries should I be familiar with for this role?

Familiarity with tools and libraries such as GitHub, PyTorch, TensorFlow, Visual Studio, and ONNX is important for this internship.

What is the hourly pay rate range for this intern position?

The hourly pay rate range for this role is between $44 and $63 USD, depending on various factors such as location and individual qualifications.

Are there opportunities for communication and collaboration across different engineering disciplines?

Yes, effective verbal and written communication skills are crucial, as you will need to collaborate with various engineering audiences, including software, hardware, architecture, and leadership teams.

Will this position be conducted in an office environment?

Yes, this position will be performed in an office setting, requiring the incumbent to sit and stand at a desk and operate standard office equipment.

Does the company provide accommodations for individuals with disabilities?

Yes, Samsung Research America is committed to complying with all laws related to the employment of qualified individuals with disabilities and provides reasonable accommodations as needed during the selection process.

image

Samsung Research America (SRA)

Technology
Industry
1001-5000
Employees

Mission & Purpose

Founded in October 1988, Samsung Research America (SRA) builds upon Samsung’s 40-year history in the Bay Area. Headquartered in Silicon Valley, with offices across the United States and Canada, SRA is engaged in researching emerging technology to create new businesses and developing core technology to enhance the competitiveness of Samsung products. We also play a key role in providing the infrastructure to support Samsung’s open innovation and university collaboration activities.