**Circom** is a domain specific language (DSL) central to **modern zero-knowledge proof systems**, powering cryptographic protocols and privacy-preserving computations. Writing robust Circom code requires both logical reasoning and syntactic precision, something current LLMs often struggle with.

**Your task is to delve into LLMs and ML engineering for code generation - find a way to get models to generate correct Circom code!**

Your submitted notebook will then run on kaggle and your model will be evaluated against 50 problems on Circom, and it must generate the code for implementing each of them. 

If the circom code generated by your model passes all the test cases for a problem, you get one added to your score. And so, there are 50 total problems to be grabbed. The best model wins. 

**The higher you score the more points you get : )**

Originally a part of Silicon Maze 2025, now open for all : )