- Software Solutions with Generative AI- RAG and Multi-Agent AI Systems- Voicebot and Chatbot- Natural Language Processing (NLP)- MLOps- Robotic Process Automation (RPA)

• Chatbot and Voice bot development• Sentiment analysis, Text classification• Machine Learning model deployment• API & Library: Pytorch, Hugging Face (transformers), Flask, Docker, Pandas, Numpy and others• Language: Python• Applications: Visual Studio Code

Title: Advanced Component Connection Extraction from Handwritten Circuit Diagram Images.Contribution:• Extension of the CGHD dataset (public dataset for hand-drawn circuit diagrams) by adding rotation annotation, circuit diagrams, segmentation maps and inspecting the bounding box for overlaps. ( DATASET LINK: https://www.kaggle.com/datasets/johannesbayer/cghd1152?select=segmentation.py )• Development of algorithms for data conversion and refinement.• Development of… Show more Title: Advanced Component Connection Extraction from Handwritten Circuit Diagram Images.Contribution:• Extension of the CGHD dataset (public dataset for hand-drawn circuit diagrams) by adding rotation annotation, circuit diagrams, segmentation maps and inspecting the bounding box for overlaps. ( DATASET LINK: https://www.kaggle.com/datasets/johannesbayer/cghd1152?select=segmentation.py )• Development of algorithms for data conversion and refinement.• Development of algorithms to automate the data annotation (rotation and connector types).• Trained UNET Model for Semantic Segmentation• Trained Mask-RCNN models for instance segmentation and keypoint detection.• Graph Extraction using AI ModelAbstract:The thesis discusses the need for digitizing hand-drawn electrical circuits using machine learning technologies to save time and reduce errors. An instance segmentation model is trained to extract the connection between different circuit symbols and identify them. The thesis also investigates the possibility of automatically transforming data from one format to another while maintaining data quality. The transformed dataset is used to train semantic segmentation, instance segmentation, and keypoint detection models, which are evaluated with various metrics. Automatic data annotation is explored, and a Convolutional neural networks (CNN) model is trained to demonstrate automatic rotation and connector-type annotation. Overall, this thesis contributes to the digitalization process of handwritten circuit recognition using machine learning and extending the dataset. The results of this research intend to improve the integration of analogue circuit drawings into digital systems. Show less

Responsibilities:• Generating data using simulation tools and preprocessing the data (approx. 4 M. samples)• Researching and creating models (DNN, CNN, Decision Tree, Gradient Boosting Tree, SVM etc)• Pruning, Model quantization (float32 to int8) and statistical analysis of the created model• Analysis of the hardware complexity for the created model• Research about hardware implementation of the created modelTools and Language:• API & Library: Tensorflow Keras… Show more Responsibilities:• Generating data using simulation tools and preprocessing the data (approx. 4 M. samples)• Researching and creating models (DNN, CNN, Decision Tree, Gradient Boosting Tree, SVM etc)• Pruning, Model quantization (float32 to int8) and statistical analysis of the created model• Analysis of the hardware complexity for the created model• Research about hardware implementation of the created modelTools and Language:• API & Library: Tensorflow Keras, scikit-learn, Pytorch, Tensorflow Lite, Pandas, Numpy and others• Language: Python• Applications: Visual Studio Code, Jupyter Notebook Show less

• Design, Implementation, Synthesis and Test of Embedded Systems• Tools: Xilinx Vivado HLs, git• Language used: VHDL and assembly• Developing hardware and software for the target platform (RISC CPU, called LT16x32) in a Linux environment and testing the designs.• Lab Project: Designing clients in a distributed system in which the communicationbetween clients is established based on the CAN bus protocol

• Property checking of digital systems by using property checking language SVA• Tool: OneSpin 360 DV• Completeness check of property set by using case split test, determination test, reset test and successor test

• Lead 25 employees to deliver outstanding output• Secure investment of 30,000 € for the company• Lead the IT team and provide them with Functional requirements document• Developing strategies for the company.• Develop the business from scratch and achieve 600+ loyal customers