Welcome to my GitHub profile! I'm a passionate student pursuing an Integrated Dual Degree in Mathematics and Computing at IIT (BHU), Varanasi, with a strong interest in AI/ML research, computer systems, and algorithm design.
- Built a dataset of 3.4M+ images, with reasoning generated for 1M+ fake images using Gemini-1.5-Pro.
- Designed a novel architecture leveraging Variational AutoEncoders with 99% accuracy on similar latent-space images and 75%+ accuracy on unseen datasets (Adobe Firefly).
- Fine-tuned Qwen2-VL with Class Activation Mapping for image reasoning.
- Tools: PyTorch, PyTorch Lightning, Gemini API.
- Developed an architecture for GIF Visual Question Answering using qformer from BLIP-2 and Llama3.2-1b.
- Explored Vision-Language Models (VLMs) and their fine-tuning potential on GIF data.
- Tools: PyTorch.
- Trained a neural network to solve Ordinary Differential Equations under the guidance of Dr. Santwana Mukhopadhyaya.
- Explored Graph-based ML and how neural networks learn functions.
- Tools: PyTorch.
- Languages: C++, C, Python, JavaScript
- Frameworks & Tools: PyTorch, TensorFlow, LangChain, Cirq, Qiskit, MERN Web Development
- Understanding the World’s Museums through Vision-Language Reasoning
Curated a large-scale dataset of 65M images and 200M question-answer pairs for benchmarking vision-language models across visual question answering tasks.
- Portfolio: rasesh2005.github.io
- LinkedIn: linkedin.com/in/rasesh-shetty
- GitHub: github.com/Rasesh2005
- CodeChef: codechef.com/users/rasesh_shetty
- Codeforces: codeforces.com/profile/ReaperScythe21
Feel free to explore my repositories and connect!