Shuhul Handoo

Organized a research problem of re-identification problem via keypoint-matching algorithm using th state-of-the-art model, SuperGLue. Managed and pre-processed datasets for training purposes via developing efficient dataloader.

Developed web application in FastAPI, for question-answering tasks for pdf files using pipelines through Langchain and HuggingFace. Conducted thorough testing of API endpoints using Postman. Extended the above project for websites and html, consisting of web-sscraping using BeautifulSoup (Python). Developed a chrome extension (NodeJs) for replying the tweets through memes generated by OpenAI APIs and Stable Diffusion

Labelling of the food datasets into respective classes using LabelStudio and developed a CNN-based pretrained models (ResNet50 and VGG19) for food-images classification, achieving an accuracy of 74%