- MS in Computer Science - Rochester Institute of Technology
- BTech in Electronics and Communication - Manipal Institute of Technology
- Financial fraud detection through a retrieval-augmented classification framework
- Used pre-trained SAINT encoder with custom retrieval
- 2.209% improvement over SAINT on benchmark datasets
- Tech Stack: Python, Deep Learning, SAINT Architecture, RAGs
- Preference prediction system for the LMSYS Chatbot Arena competition to predict human preference scores between pairs of LLM responses.
- QLoRA with 4-bit quantization
- Hybrid architecture with Gemma-2B and Llama-3B
- Tech Stack: PyTorch, LLMs, QLoRA, GPU Computing
- Lenet5 implementation with custom CUDA kernels for convolution operations
- 16x speedup in convolution operations
- 95% accuracy on MNIST dataset
- Tech Stack: C++, CUDA, CNN Architecture
- Automated code comment updates based on code changes
- Hybrid architecture combining CUP with HebCUP
- 10% accuracy improvement through sub-token classification
- Tech Stack: NLP, Machine Learning, Python