ML-Research

Assorted projects that I made to learn about AI, ML, and LLMs. My own personal experimentation.

1-finetune_llama2_user_reviews

Given a list of user actions on an ecommerce website (click sizing tool) etc, this fine tuned version of llama3 LLM will generate potential user reviews. For example, for a user who previously clicked on the sizing tool, the llm will generate 'I loved the sizing tool becuase it helped me find the perfect size! The fit of the shirt was amazing.' This is useful technology because it can help boost reviews (by giving suggestions to users) with the use of AI. Read More: https://github.com/shreybirmiwal/finetuned-llama2-user_reviews

2-quantization-demo

Quantization is done to reduce the # of bits and therfor the reduce the complexity of computation in the weights of a neural network. I experimented and implemented my own quantization demo. Read more at the blog: https://shreybirmiwal.vercel.app/blog/0

3-reinforcement-learning and 4-dino_game_rl

Experimenting with reinforcement learning

7-micrograd and 8-makemore

Working through Karpathy's NN 0-hero course. Built micrograd, a autogradient generator for neural networks, and working through makemore, various implementations of basic LLMs.