Aiml Data Scientist - Bengaluru, Karnataka, India - Skillety Technologies

Skillety Technologies

Verified Company

Bengaluru, Karnataka, India

3 weeks ago

Posted by:

Deepika Kaur

beBee Recuiter

Description

Job Description Summary

1) Strong 'C/C++/python' Programming skills. Good knowledge of data structures & algorithms
2) Understanding of computer architecture (ARM processor architecture, cache, memory, bandwidth, parallel and distributed processing concepts)
3) Experience with common ML frameworks like caffe, TensorFlow, pytorch, TensorflowLite, ONNX.
4) In-depth understanding on pre-processing pre-trained models to optimize the inference.
5) In-depth understanding in optimization techniques like fusing of multiple operations, kernel optimization, compiler optimization techniques.
6) Strong trouble shooting and problem solving skills.
7) Understanding of Android ecosystem & hands on with Android NDK based native module development.

o

o

1)Designing and implementing support for various vendor frameworks in existing acceleration platform to leverage CPU/GPU/NPU & DSP

2) Understanding the open source/vendor compiler way of building the network in NPU understandable format and customizing the same for in existing acceleration platform usage with optimization techniques like pre-processing, pruning, quantizing the network etc.

3)Developing custom tool to generate a network capable of being understood by both vendor as well as in-house engines.

4) Utilizing the open source model optimization techniques and enabling the lighter model to Flagship and Mid-Mass models

4)Optimizing the runtime framework for DNN models using various optimization techniques on all possible compute units.

5) Commercialization support for DNN models on all Samsung mobile devices [innovative/premiuM/Flagship].

6) Propose the alternate layers in DNN architecture to achieve the best accuracy and good performance.

o

o

Additional Job Description

o

1)Prior experience in developing DNN model or optimizing the model for different ML frameworks.

2) Prior understanding of Tensorflow-lite & Pytorch mobile runtime library.

3) Prior experience of coming up with alternate layers based on the supporting ops for each compute unit for all ML frameworks.

4)Prior experience with understanding the Opensource frameworks and the process of converting a DNN model to chipset vendor specific file. Converting a model from one ML framework to other supporting framework.

5)Prior experience or understanding of Android Neural Network APIs and runtime will be an added advantage.