ML
PERFORMANCE
ENGINEER

Deep Render is the world-leading R&D team for AI-based video and image compression technologies, with 50+years of combined research experience and 20+ patents. Our team stems from world-leading research institutions such as Imperial, Oxford, UC Berkeley and McGill University with publications in top-tier journals and conferences such as CVPR, ICCV, ICML, ECCV, and NeurIPS.

We’re looking for Performance Engineers to join us to be part of a highly talented team bringing the next step-change in compression technology to billions of users, creating enormous global impact and value.

We're looking for engineers who will help us deliver AI-based compression to end-users by porting our codec from GPU-based systems to (primarily) mobile platforms with NPUs and (secondarily) mid-range GPU/CPU systems, thus going from research to production. You’ll enjoy working with low-level code and are comfortable with programming across multiple platforms.

The ideal candidate will have a deep understanding of optimisation methodologies to reduce runtime and memory footprint, preferably for neural networks; and/or experience implementing high-performance entropy coding algorithms such as Huffman Coding, Arithmetic Coding, Range Coding, or Asymmetric numeral systems. The ideal candidate will have some experience with taking algorithms from research to deployment.

 
 
research-item-1

Responsibilities

  • Work in a team to port ML research algorithms to edge devices with an initial focus on smartphones (Android, iOS)
  • Profile various algorithms to analyse performance and identify any bottlenecks. Profiling includes data loading, data movement, data caching, operation count, execution chipset, warm-up latency and others
  • Implement solutions to the identified bottlenecks
  • Implement a high-performance entropy coding algorithm, e.g. Range Coding or Asymmetric Numeral Systems, across different hardware architectures
  • Optional: Write custom operations using the low-level API for Android (OpenGL ES) and iOS (Metal) systems
  • Optional: Apply standard neural network runtime optimisation methods such as pruning, low-bit quantisation, architecture tuning, batching and others

Requirements

  • At a minimum, a Bachelor's degree in computer science or related field (Mathematics, Physics, Engineering)
  • At a minimum, 3-5 years of experience in performance optimisation
  • One of the following: Either some formal training in machine learning (understanding PyTorch and/or Tensorflow) or some formal training in entropy coding methods (understanding Range Coding or similar algorithms).
  • Formal training could come through education, work experience and/or extensive private projects.
  • Expertise in Python (Expert) and C++ (Semi-Expert)
  • Some experience with optimisation techniques. Examples include SIMD (SSE, AVX), vectorisation, loop dependencies, multithreading, multi-processor usage, and tensor cores

Preferred skills

  • A strong machine learning background
  • Significant experience with ML-programming in either Android and IOS: Android Studio, XCode, Google ML, Core ML. Knowledge of the development stack for Android and iOS
  • Experience with Android NNAPI and or other Android-based NPU SDKs (Exynos, Hexagon HiSilicon)

Benefits

  • If you do not have the right to work in the UK, we can sponsor your visa
  • The newest and best equipment: Notebooks/Macbooks, 4K dual screens, mechanical keyboards, drawing pads, headphones and standing desks
  • A comprehensive private health insurance plan by AXA
  • An amazing WeWork office overlooking Tower Bridge and the Tower of London
  • Cycle to work scheme
  • Stock options; taking part in the upside potential of Deep Render
image

Join the
Compression
revolution

We look forward to receiving your application for the position of ML PERFORMANCE ENGINEER