My Learning Journey in AI, Deep Learning, Research, and Open Source

Hey everyone! 👋

I'm a second-year Computer Science student with a growing interest in Artificial Intelligence, Deep Learning, and research. I believe the best way to learn is not just by consuming information but also by experimenting, building things, asking questions, and sharing what you discover along the way.

I wanted to use this space to document my learning journey, share interesting things I come across, and connect with people working in similar areas.

Here's what I've been up to lately.

1. My Journey at MRM Research (AI)

I've successfully completed my task phase at MRM Research (AI) and am now officially part of the team!

During this phase, I've explored several areas of Deep Learning, including:

• Transformers: Understanding the architecture behind modern language models.
• GANs: Exploring Generative Adversarial Networks and their underlying concepts.
• LLMs: Learning about Large Language Models and the developments happening in the field.
• Deep Learning: Building a broader understanding of different architectures, techniques, and ongoing research directions.

It's been a great learning experience, and I'm excited to continue learning and contributing as part of the team.

2. Exploring Mechanistic Interpretability

Apart from my work with the research team, I'm also working on my own research paper.

My current research interest lies in Mechanistic Interpretability, particularly exploring how Topological Data Analysis (TDA) can help us understand neural network representations.

I'm investigating the use of Ball Mapper to analyse Sparse Autoencoder (SAE) feature activation spaces.

The idea is to explore whether topological structures and relationships within these feature spaces can help us better understand model behaviour and identify patterns associated with potential misalignment.

There's still a lot to explore, experiment with, and understand. I'll be sharing more about the methodology, implementation, challenges, and findings as the research progresses.

3. Preparing for Google Summer of Code (GSoC)

Another goal I'm actively working towards is Google Summer of Code (GSoC).

I'm spending time improving my programming skills, understanding open-source development workflows, exploring potential organizations and projects, and learning how to make meaningful contributions to existing codebases.

I want to become a better open-source contributor and gain experience working on real-world software projects.

4. Strengthening My Machine Learning Fundamentals

I'm also currently reading Hands-On Machine Learning with Scikit-Learn, and PyTorch.

My aim is to strengthen my understanding of machine learning fundamentals while getting hands-on experience implementing different algorithms and techniques.

I believe having a strong grasp of the fundamentals is just as important as keeping up with the latest developments in AI.

5. What's Next?

I want to make this space a public learning journal where I document my progress rather than just showcase achievements.

Going forward, I plan to post regularly about:

• Concepts and architectures I'm learning in AI and Deep Learning.
• Interesting research papers and my interpretations of them.
• Experiments with SAEs, Ball Mapper, and Mechanistic Interpretability.
• My GSoC preparation and open-source contributions.
• Projects, implementations, technical challenges, and lessons learned.

I don't claim to be an expert in any of these areas. I'm here to learn, experiment, make mistakes, and improve.

Hopefully, documenting this journey will help me learn better while connecting with people who share similar interests.

Here's to learning in public, building consistently, and sharing everything along the way! 🚀

Feel free to connect, share suggestions, or discuss anything related to AI, Deep Learning, and open source. I'd love to learn from this community.