Everyone has a story.
Here's mine, so far.
I enjoy chasing questions that don't have obvious answers. Most of the time that means building software. Occasionally it means writing about what I learned along the way.
One question that kept bothering me was:
If LLMs are good at understanding scientific literature, why are they still so bad at exploring ideas that aren't already obvious?
That question became SciNets.
The idea was simple instead of asking an LLM to repeat what was already written, could we build a system that explores connections hidden across different papers and surfaces ideas worth investigating?
Building it taught me far more than I expected, and eventually led me to write a paper about what worked, what didn't, and what I learned along the way.
Explore the project
• Website: https://scinets.in
• Research Paper: https://arxiv.org/abs/2601.09727
• Source Code: https://github.com/100hard/SciNets
- AI Infrastructure
- Backend Engineering
- Distributed Systems