• Researched state-of-the-art NLP methods and algorithms• Designed a production-ready algorithm for semistructured text classification from scratch ( using a combination of different word embeddings, CNNs and attention-based models)• Developed said algorithm in Tensorflow and PyTorch(lightning) in a highly maintainable manner• Tracked experiments (hyperparameters, metrics, ...) and deployed models as a REST API using MLFlowframework• Developed custom Tensorflow and Pytorch layers/modules• Designed and implemented novel fully end-to-end attention-based model for segmentation of data tables

- worked on adding static DHCPv4 leases to systemd open source project- designed and build a highly general end extendable system for remote management and monitoring of a network devices using Berkeley Packet Filter (eBPF) technology and bleeding-edge tools.- helped lead the team of interns with building edge-based machine learning techniques for networking applications

- Researched and adopted state-of-the-art machine learning algorithms- Co-designed and developed a complete machine learning pipeline for IoT combining edge computing and cloud computing- Co-designed and built pose estimation pipeline and object counting (Python, OpenPose, LSTM, PyTorch)- Written library for GPU jpeg compression using MMAL API- Implemented background subtraction algorithm using computer vision principles (C++, Python)- Mentored and lead four students during their internships- Introduced git workflow, handled code review and merge/pull requests- Co-designed and developed modern C++ application on an embedded IoT device (GUI, image processing, database, threads, ...)- Implemented several sensor libraries (I2C, SPI, Bluetooth)- Developed modern Python GUI application- Experimented with popular machine learning algorithms like YOLO and SSD (Python, TensorFlow and PyTorch)- Built C++ cloud client application (MQTT, Protobuf, REST)

- Annotated face images for emotion detection