This repository serves to explore different machine learning tools, techniques and research. It will consist of Jupyter Notebooks used to try out new developments. It will have a focus on multimodal models. A list of ideas follows:

• Research on Causality
• Study and implement CNNs, RNNs, and Transformers.
• Explore multimodal learning with image and text data.
• Pretrain models using self-supervised learning techniques.
• Pretrain and finetune LLMs (e.g., GPT, BERT).
• Study optimizers, scaling laws, and distributed training.
• Work with vision models like ViT and ResNet.
• Fine-tune LLMs for NLP tasks (e.g., summarization).
• Train speech models like Wav2Vec and Whisper.
• Fine-tune multimodal architectures like CLIP and Flamingo.
• Study reinforcement learning and RLHF.
• Work with GANs, VAEs, and diffusion models.
• Implement RAG pipelines with LLMs and knowledge graphs.
• Experiment with cross-modal alignment techniques.
• Fine-tune pre-existing multimodal models.
• Train large foundation models using parameter-efficient fine-tuning.
• Build multimodal applications (e.g., video QA, chatbot).
• Study scaling techniques like model parallelism and Megatron-LM.
• Create and curate custom multimodal datasets.
• Deploy multimodal models using cloud services and APIs.