Edge ML Engineer
As a Machine Learning Engineer, you will join a team that takes models and adapts them to run on low-power edge devices, using a variety of techniques such as pruning, quantization, distillation and more.
In light of the development of a new field in Webiks, Machine Learning at the Edge, Webiks is looking for a ML Engineer, who will develop software and algorithmic aspects of adapting ML models to run on low-power Edge devices.
- Explore Low-Power hardware, suitable for computer vision missions (e.g. TPU and Neuromorphic processors), and for other missions such as NLP and STT.
- 'Slim' large neural networks using techniques such as Pruning, Quantization,Clustering and Knowledge Distillation, with minimal accuracy loss and significant improvement in latency and energy consumption.
- Implement algorithms for early stopping of inference, after reaching a sufficient level of confidence in the predictions, to allow a reduction of running time and improved energy efficiency (such as BranchyNet).
- Develop and implement algorithms for adaptive calibration of the inference “operating point” at the Edge, calibrating tradeoffs such as Precision versus Latency.
- Use dedicated toolchains that provide complete pipelines for fitting and running models on edge devices (such as Edge Impulse)
- Work with dedicated frameworks for Edge-ML, such as TensorFlow Lite and Tensor-RT.
- Write low-level code in C/C++.
- In-depth familiarity with Deep Learning models, with an emphasis on computer vision.
- Familiarity with running models Low-Power devices (ML at the Edge).
- Familiarity with frameworks such as TensorFlow Lite or Tensor-RT - an Advantage.
- Bachelor’s degree in computer science, engineering or other relevant profession - Required (Master’s degree - an Advantage).