My Projects
Exploring technology and building solutions
🧠 fusedNeural: From-Scratch Neural Network in C
C
Numerics
Performance
Quantization
A low-level, from-scratch neural network written in C with an emphasis on performance for resource-constrained environments. The project implements a single-layer network with forward and backward passes, an MSE loss function, kernel fusion for improved cache locality, and post-training quantization to reduce memory footprint while maintaining accuracy. It includes a small backpropagation engine and utilities for tensor math.
- Forward and backward passes with backpropagation (MSE loss)
- Kernel fusion to minimize memory bandwidth and improve throughput
- Post-training quantization of tensors and operations
- Modular components: tensors, ops, quantized ops, loss functions
Core primitives complete; exploring multi-layer extensions and benchmarking