Convolutional neural network from scratch
from_scratch
deep_learning
first_principle
cnn
Tech Stack
Rust
Docker
Terraform
ndarray
Description
Full from-scratch CNN implementation in Rust — no ML frameworks, no black-box abstractions. Forward pass, backpropagation, and gradient computation implemented manually based on a paper, demonstrating in-depth understanding of the underlying mathematics rather than framework-level API usage.
Chose Rust to deepen systems programming skills alongside the ML math.
Read more in my substack article.