Project summary

I am a final-year university student in Kosovo aiming to transition into AI safety research, with a focus on mechanistic interpretability.

This project will allow me to:

• Complete my remaining degree obligations

• Dedicate 3–4 months of full-time study to AI safety

• Build practical, open-source interpretability projects

What are this project's goals? How will you achieve them?

Goals:

• Gain a solid foundation in mechanistic interpretability

• Build 2 practical projects analyzing neural network behavior

• Become capable of contributing to AI safety research communities

Approach:

• Study core materials (e.g. transformer interpretability, circuits)

• Reproduce key experiments from papers

• Build small tools to visualize and analyze model behavior

• Share all work publicly on GitHub
  

How will this funding be used?

The funding (approx. $1500–2000) will be used to:

• Cover remaining university tuition obligations

• Cover basic living expenses for 3–4 months

• Enable full-time focus (no part-time work required)

This uninterrupted time is critical for deep technical learning.

Who is on your team? What's your track record on similar projects?

I am currently an individual contributor.

Background:

• Ongoing university studies (near completion)

• Completed online courses in AI/ML (e.g. Andrej Karpathy materials, Arena program)

• Strong motivation to specialize in AI safety

While I do not yet have major research outputs, this project is designed to produce my first serious contributions.

Initial progress:

I have already implemented a full automatic differentiation engine (micrograd-style) from scratch in Python.

GitHub: https://github.com/mejdisadriu/micrograd-from-scratch

This demonstrates early-stage progress toward understanding neural networks at a mechanistic level.

What are the most likely causes and outcomes if this project fails?

Risks:

• Underestimating difficulty of interpretability research

• Lack of prior research experience

• Possible slower-than-expected progress

Mitigation:

• Focus on small, achievable projects first

• Follow structured curricula and existing research directions

• Engage with online communities for feedback

Even in the worst case, the outcome will still be:

• Stronger ML foundation

• Public portfolio of work

• Increased employability in AI-related roles
  

How much money have you raised in the last 12 months, and from where?

I have not raised any external funding in the past 12 months.