• Selected as a Summer Student at CERN under the Patatrack team.• Working on the CMSSW profiling and optimization for the Event Reconstruction Software for the Large Hadron Collider.

• Trained and deployed image segmentation models, including DeepLabV3 and U2Net, for image editing application.• Actively working on code optimization and development to decrease production downtime and improve user experience.• Implemented and improved a RAG pipeline for document-based question answering with HuggingFace and PineconeDB.• Deployed QwenVL for image question answering application, integrating Amazon S3 and Redis for efficient computation.

• Worked on integration of Machine Learning models in FastAPI infrastructure and containerized deployments for seamless integrations.• Worked on GitHub Actions for continuous integration of workflows for easier collaborations.• Focused on optimizing high-traffic models, resulting in a two-fold reduction in inference time and server costs.• Actively working on code optimization and development to decrease production downtime and improve user experience.• Developed and deployed… Show more • Worked on integration of Machine Learning models in FastAPI infrastructure and containerized deployments for seamless integrations.• Worked on GitHub Actions for continuous integration of workflows for easier collaborations.• Focused on optimizing high-traffic models, resulting in a two-fold reduction in inference time and server costs.• Actively working on code optimization and development to decrease production downtime and improve user experience.• Developed and deployed Machine Learning models for Real-time Sketch generations, CLIP-model encoding, and Generative Background changers using Computer Vision. Show less

• Utilized simulations to design transitional spaces, considering visual comfort analysis by incorporating variations in visual field cones and measuring points within the transitional area.• Engaged in the creation of an interpolation model designed to predict daylight luminance within transitional spaces throughout the entire year.• Leveraging machine learning algorithms to interpolate visual comfort metrics within shoe-box simulations.

• Trained and enhanced GAN-based image super-resolution models to significantly elevate user satisfaction.• Optimized architectures using ONNX and TensorRT conversions and deployed compiled graphs for faster inference.• Researched, trained, and developed Stable Diffusion applications including ImagineAI and AvatarMe.• Finetuned models to enhance image quality, rivaling Canva and MidJourney on an international scale.• Introduced a range of innovative app functionalities… Show more • Trained and enhanced GAN-based image super-resolution models to significantly elevate user satisfaction.• Optimized architectures using ONNX and TensorRT conversions and deployed compiled graphs for faster inference.• Researched, trained, and developed Stable Diffusion applications including ImagineAI and AvatarMe.• Finetuned models to enhance image quality, rivaling Canva and MidJourney on an international scale.• Introduced a range of innovative app functionalities, encompassing Image Generation, Image Editing, and Control-net. Show less

• Worked on both frontend and backend technologies, gaining experience with ReactJS and NodeJS.• Debugged major sections of the website that allowed vendors to gain access to their profiles. Automated a previously manual task.• Integrated advertisements, multiple communication channels, and notifications to substantially increase customer interaction and increase revenue streams.• Added a major booking functionality that allowed customers to book vendors through the website.

• Implemented classification, segmentation, and object detection models including ResNet, UNet and RCNN in collaboration with PH.D. students for Computer Vision projects.• Scraped, annotated and compiled a novel dataset of over ten thousand images for oriented object detection task.• Engaged in the expansion of object detection capabilities, aiming to automate the process of identifying target objects and generating oriented bounding boxes around them, using FasterRCNN and YoLo models.

• Collaborated with Ph.D. research scholars to collaborate on the development of Natural Language Processing architectures and methodologies, to implement a AI-based Search Engine.• Developed a Semantic Search Engine tailored for a specialized Pakistani Legal Dataset, employing machine learning models to facilitate rapid query searches based on contextual cues.• Fine-tuned Transformer-based architectures, including BERT, using a semi-supervised approach to enhance the performance of the… Show more • Collaborated with Ph.D. research scholars to collaborate on the development of Natural Language Processing architectures and methodologies, to implement a AI-based Search Engine.• Developed a Semantic Search Engine tailored for a specialized Pakistani Legal Dataset, employing machine learning models to facilitate rapid query searches based on contextual cues.• Fine-tuned Transformer-based architectures, including BERT, using a semi-supervised approach to enhance the performance of the search engine. Show less