• 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗲𝗱 𝗗𝗮𝗶𝗹𝘆 𝗥𝗲𝗽𝗼𝗿𝘁 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 𝘄𝗶𝘁𝗵 𝗔𝘇𝘂𝗿𝗲 𝗙𝘂𝗻𝗰𝘁𝗶𝗼𝗻𝘀: Implemented a scheduled automated report generation system using Azure Functions, fetching category-based data from Azure Blob Storage to provide daily analytics in a predefined report template.

𝗣𝗿𝗼𝗱𝘂𝗰𝘁: 𝗥𝗮𝗱𝗮𝗿𝗜𝗺𝗮𝗴𝗲𝗿• Conducted experiments comparing YOLO, RetinaNet, and Vision Transformers (ViT) for object detection on industrial 3D imaging systems based on radar technology.• Focused on handling food products, such as eggs and chocolate boxes, ensuring accurate detection to maintain product quality and reduce damage.• Enhanced quality assurance by improving detection of anomalies, ensuring safety and compliance in the packaging process.• Performed FMCW radar signal processing techniques as a pre-processing step for model training.

• Experimented to analyze the material emissivity of ps16 3D printed material, exploring variations in shapes and forms.• Investigated the temperature dependence of ps16 material emissivity by calculating it with a reference object in a heating cabin set at 150 degrees Celsius.• Utilized Python for data analysis, visualization, and result presentation.• Successfully executed experiments and presented compelling visualizations to the team.• Contributed valuable insights into the thermal properties of ps16 material, enhancing our understanding of its behavior under different conditions. Tools & Technologies: Python, FLIR IR camera calibration

• Worked on interpretable correlation analysis between model performance and input features for automated dataset generation pipeline. Tools & Technologies: Python, Django

• Developed an innovative image similarity matching deep learning model to track retail products seamlessly from warehouse inventory to shopping shelf placement.• Implemented an anomaly detection model for identifying irregularities in printed text on retail products during the manufacturing process on a conveyor belt, ensuring quality control.• Worked on Debian package development for deep learning packages, facilitating GPU access on the Jetson Nano board from within a Docker container.Tools & Technologies: Tensorflow, Python, MLflow, Docker, FastAPI, Jetson Nano.

• Worked with Tagir Fabarisov under Dr. Andrey Morozov's Team. • Development of lightweight Deep Learning based fault detector for the exoskeleton. • Converted PyTorch model to TF-Lite for low latency hardware inference.Tools & Technologies: Raspberry pi 3, PyTorch, Matlab Simulink, TFLite

• Worked on a research project under Dr. Christian A. Mahringer and Dr. Simone MayerTools & Technologies: R Language, Bibexcel, Gephi

• 𝗟𝗼𝗰𝗮𝘁𝗶𝗼𝗻-𝗧𝗮𝗿𝗴𝗲𝘁𝗲𝗱 𝗡𝗲𝘄𝘀 𝗥𝗲𝗰𝗼𝗺𝗺𝗲𝗻𝗱𝗮𝘁𝗶𝗼𝗻𝘀: Implemented a Named Entity Recognition (NER) model with a prior probability distribution and achieved an 8% improvement in click-through rate.• 𝗔/𝗕 𝗧𝗲𝘀𝘁𝗶𝗻𝗴 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝘆 𝗳𝗼𝗿 𝗥𝗲𝗰𝗼𝗺𝗺𝗲𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗔𝗹𝗴𝗼𝗿𝗶𝘁𝗵𝗺: Designed and developed an A/B testing strategy, achieved statistical significance, and validated performance improvements.• 𝗔𝗻𝗮𝗹𝘆𝘁𝗶𝗰𝘀 𝗗𝗮𝘀𝗵𝗯𝗼𝗮𝗿𝗱 𝗳𝗼𝗿 𝗖𝗹𝗶𝗰𝗸-𝗧𝗵𝗿𝗼𝘂𝗴𝗵 𝗥𝗮𝘁𝗲 𝗧𝗿𝗮𝗰𝗸𝗶𝗻𝗴: Developed dashboard in Redash to monitor the click-through rates of articles across multiple publishers. Leveraged Apache Kafka for event data and Apache Druid for distributed data storage.• 𝗞𝗣𝗜 𝗖𝗮𝗹𝗰𝘂𝗹𝗮𝘁𝗶𝗼𝗻 𝘄𝗶𝘁𝗵 𝗣𝘆𝗦𝗽𝗮𝗿𝗸: Developed PySpark jobs to determine average articles per user and average time spent per article across mobile app and web platforms daily.• 𝗦𝗰𝗮𝗹𝗮𝗯𝗹𝗲 𝗣𝗿𝗼𝗯𝗮𝗯𝗶𝗹𝗶𝘀𝘁𝗶𝗰 𝗔𝗹𝗴𝗼𝗿𝗶𝘁𝗵𝗺: Developed a probabilistic algorithm using Python and the WebEngage platform to send recommendations via push notifications, targeting 1,000 notifications per second, which increased user interaction by 10%.• 𝗢𝗽𝗲𝗻-𝗦𝗼𝘂𝗿𝗰𝗲 𝗖𝗼𝗻𝘁𝗿𝗶𝗯𝘂𝘁𝗶𝗼𝗻: Contributed to the SimDB project (Semantic search using Faiss) by adding a feature to preprocess Kafka events to significantly increase throughput.

• 𝗘𝗻𝗵𝗮𝗻𝗰𝗲𝗱 𝗢𝗖𝗥 𝗮𝗻𝗱 𝗛𝗮𝗻𝗱𝘄𝗿𝗶𝘁𝗶𝗻𝗴 𝗥𝗲𝗰𝗼𝗴𝗻𝗶𝘁𝗶𝗼𝗻: Developed a Super Resolution Model that improved Optical Character Recognition (OCR) and Handwriting Text Recognition accuracy by 15%.• 𝗖𝗼𝗻𝘁𝗲𝘅𝘁-𝗔𝘄𝗮𝗿𝗲 𝗦𝗽𝗲𝗹𝗹𝗶𝗻𝗴 𝗖𝗼𝗿𝗿𝗲𝗰𝘁𝗶𝗼𝗻: Implemented a word2vec model for spelling correction that leveraged text context, resulting in a 10% improvement in OCR accuracy.• 𝗜𝗻𝘁𝗲𝗿𝗻𝗮𝗹 𝗗𝗮𝘁𝗮 𝗔𝗻𝗻𝗼𝘁𝗮𝘁𝗶𝗼𝗻 𝗧𝗼𝗼𝗹: Created an efficient data annotation tool using the Python Flask framework, with annotated data seamlessly stored in MongoDB.• 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗲𝗱 𝗦𝘂𝗿𝘃𝗲𝘆 𝗗𝗮𝘁𝗮 𝗘𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻: Automated the extraction of information from survey forms using kernel-based image processing techniques with OpenCV-python, reducing human intervention. Achieved processing times of 60 ms per document, averaging 6 pages each.• 𝗘-𝗰𝗼𝗺𝗺𝗲𝗿𝗰𝗲 𝗖𝗵𝗮𝘁𝗯𝗼𝘁𝘀 𝗳𝗼𝗿 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 𝗥𝗲𝗰𝗼𝗺𝗺𝗲𝗻𝗱𝗮𝘁𝗶𝗼𝗻𝘀: Developed intelligent chatbots for an e-commerce website, enhancing product recommendations and increasing product visibility.• 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗲𝗱 𝗔𝗻𝗮𝗹𝘆𝘁𝗶𝗰𝘀 𝗥𝗲𝗽𝗼𝗿𝘁𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗔𝗪𝗦: Developed an AWS Lambda function to generate automated analytics reports, triggered by data uploads to AWS S3 and integrated with the SNS service for notifications.