PROJECTS 💻
> I believe that the best way to learn new stuff is by making a cool project out of it. These are some things that I've made while on this learning journey.
> SONG RECOMMENDATION SYSTEM
> Apart from making regressors and classifiers using datasets like california housing, or the iris dataset, this was my first proper experience making a machine learning model with an online dataset.
> I found an online dataset that had around 100,000 songs in a .csv file. All those songs were classified into different subgenres with repetition of a few songs in places. I took that dataset, cleaned it with basic EDA techniques and then used AnnoyIndex by spotify to make a mini song rec system myself.
> ASL DETECTION USING YOLOv5
> This project was a result of my interest in computer vision integrated with AI.
> Using a dateset of around 2000 images from Roboflow and then the yolov5 model trained on a modest gpu, I was able to make the model detect some 18 out of the 26 alphabets in the English language.
> NEURAL NETWORK FROM SCATCH
> This is not a project, but something that im currently working on to learn more about how neural networks work.
> With my understanding of calculus, I started learning the math behind neural networks. This includes core concepts like Forward Propagation, Backpropagation, and activation functions. So, I used what i learned to implement a very simple neural network which includes a Fully Connected layer and activation functions. This was done in python from scratch usng only the numpy library.
> LOCALLY FINE TUNED GPT-2 MODEL
> This is probably the most fun I had while making a project.
> This was a random idea that came to my mind - what if I fine tuned an LLM so that it could answer questions about me, and then add it as a cool feature in my website? So I went straight to work, created a jsonl file containing 500 questions and answers about myself, and used it to fine tune a gpt-2 model. The results were good. Although I went ahead with using Groq's /Together.ai's api for this website, making that project allowed me to understand core LLM concepts like tokenization, fine tuning, etc.
> And That's it for now! If you're still reading, thank you for taking your time to go through my projects! I will be adding more stuff when I make them, so stay tuned!