07.09.2022 aktualisiert


100 % verfügbar
Machine Learning Engineer
Munich, Deutschland
Deutschland +2
MSCEinfo: Deutschland, Österreich, Schweiz
Skills
MxnetAgile MethodologieAlgorithmusAmazon Web ServicesAmazon Elastic Compute CloudData AnalysisKünstliche Neurale NetzwerkeComputer VisionAtlassian ConfluenceAtlassian JiraMicrosoft AzureUNIXC++Cloud ComputingComputerprogrammierung
- Machine learning, deep learning, AWS, algorithms, neural networks
- Programming: Python, Matlab, C++, TCL, VHDL, Keras, TensorFlow, Numpy, OpenCV, Scikit-learn, Pandas, Horovod, MXNet Gluon, PyTorch
- Kubernetes, Docker, Comet-ML, cloud, Azure (VMs), EKS, EC2, GCP, GCE, Google, GPU
- PostgreSQL, MySQL, SQLite
- version control systems, Bitbucket, Gitlab, GitHub
- CI/CD, Jira, Confluence, Agile Scrum, Unix, Linux
Erfahrung
Berufserfahrung0 - 5 Jahre
Sprachen
DeutschGrundkenntnisseEnglischverhandlungssicher
Projekthistorie
? * Designed, developed, and optimized deep learning-based algorithms for multiple objects (pedestri-½
ans, bikes, cars, trucks e.t.c.) detection and classification using Radar image data and range and velocity
vectors
* Analyzed and improved the existing object classification solutions and the training data pipeline using
tracking algorithms, auto-labeling and data augmentation techniques
* Developed pre-processing object segmentation algorithms and training procedures
* Designed and trained novel lightweight multi-modal deep neural network architectures based on
convolution networks, recursive networks, and hybrid networks
* Improved the overall test accuracy by 13% with a 35 times smaller network size
ans, bikes, cars, trucks e.t.c.) detection and classification using Radar image data and range and velocity
vectors
* Analyzed and improved the existing object classification solutions and the training data pipeline using
tracking algorithms, auto-labeling and data augmentation techniques
* Developed pre-processing object segmentation algorithms and training procedures
* Designed and trained novel lightweight multi-modal deep neural network architectures based on
convolution networks, recursive networks, and hybrid networks
* Improved the overall test accuracy by 13% with a 35 times smaller network size