● Working on detecting pulse during CPR by analysing blood-flow velocity patterns from carotid Doppler ultrasound and estimating key clinical metrics from average velocity envelopes. ● Developed deep learning models for predicting Return of Spontaneous Circulation (ROSC) using Doppler ultrasound during CPR. Classification models to detect ROSC vs other signals (completed), segmentation model to predict the average velocity envelope (in progress), and peak detection model. ● Implemented CUDA convolution kernel for training with C++ and developed models with C++ PyTorch layers and implemented TensorRT inference pipeline for edge devices. Developed MATLAB GUI for data annotation and signal preprocessing. Working on multimodal AI systems combining ultrasound and ECG signals for ROSC detection and implemented signal processing and noise reduction algorithms. ● Published research in Resuscitation Plus and MIDL conference. Supervised 3 master’s students in AI-based medical imaging projects. Projects: ● Immune System Response in Polytrauma (12 CP): used deep learning and machine learning to predict complications in polytrauma patients and identify key biomarkers. Individual project. Published a paper in the Frontiers in Immunology Journal. ● Super-Resolution of 3D DWI MRI: led the development of a UNet-based model for super-resolving diffusion-weighted MRI volumes (DICOM), resulting in a publication at the EUSIPCO conference. ● Derived Data Prediction for DWI: developed deep learning models to generate diffusion tensors and brain tracts from MRI data, along with a Qt5-based GUI for visualization. ● Hand Movement Classification: built an LSTM model in MATLAB for classifying hand movements from sensor glove data, incorporating signal processing techniques; published in MDPI Sensors. ● Personal Projects (OpenGL/Vulkan/VR): building 3D visualization tools using Vulkan, OpenGL, C++ and Android NDKs out of personal interest along with online courses (hobby coding). ● Developed Android apps using OpenGL, Vulkan SDK for 3D visualization with native C++. Developing QEMU-based Linux software.

● Improved weakly supervised deep learning model for cervical cancer detection on whole-slide images (sensitivity: 60% → 80%). Integrated artifact removal pipeline for preprocessing in large-scale medical images using SAM model and conventional computer-vision techniques. ● Reduced slide-scanning time by two minutes by improving the existing parallel processing pipeline and existing hardware-software interaction pipeline with STM32 MCU and migrated the existing CPU-based image processing workflow to GPU using CUDA and OpenCV C++ libraries. Implemented a better and faster gamma correction pipeline for WSI images using OpenCV C++ on GPU. ● Developed and maintained Qt-based cross-platform GUIs for medical imaging systems. Collaborated across software, hardware, and data teams in a startup environment.

● Developed a C++-based 3D visualization system using Vulkan SDK and OpenGL on Linux as well as Android devices. ● Integrated gesture recognition models for real-time interaction on Android-based AR headsets. Worked on edge AI deployment, real-time rendering, and human–computer interaction systems.

● Worked as an intern to detect volatile organic compounds in breath gas spectrums using ML models. Developed GUI with Qt5 for visualization of the predictions and XAI biomarkers for VOCs. ● Worked with detection and segmentation of cancerous regions and counting cancerous cells in Vulvar cancer Tissue Micro Arrays (Master’s thesis). Applied XAI techniques for identifying relevant biomarkers. Published a paper at the SPIE Medical Imaging Conference.

● Worked on CT image reconstruction and denoising using deep learning and signal processing. ● Implemented filtered back projection algorithm using Python and developed a UNet-based noise removal model.