voxium is the public name of a programmer who publishes projects through thevoxium on GitHub. The portfolio includes low-level machine-learning libraries and systems tools.[1][3]
Start with bare-lm →
Links
X profile 
GitHub profile Contributions
bare-lm is a C tensor library with automatic differentiation. Its worked example constructs a two-layer network for XOR, computes a loss, propagates gradients backward and updates parameters. Separate memory arenas distinguish persistent tensors from temporary computation, making allocation and cleanup part of the visible training workflow. The project is a small implementation a learner can inspect rather than a claim to match a full industrial framework.[3]
Tools, writing & talks
Tools
Start here
bare-lmgithub.comA minimal C tensor and automatic-differentiation library with a worked neural-network example.
Trace the XOR example from input tensors through loss calculation and parameter updates. C, memory management and basic neural-network calculus help; a compiler, make and OpenBLAS are required for the documented build.
Repository owner and developer[3]
Writing, research & talks
- Build Your Own Inference Engine: From Scratch to "7"michalpitr.substack.com
Part one of a from-scratch C++ inference engine: train a small MNIST network with a non-trivial topology, export it to ONNX, then load, graph and execute it yourself — with a working explanation of how engines differ from inference servers and which optimizations pay.
Tutorial by Michal PitrOriginal source