Abhishek Shivakumar

Systems Engineer | Founder @ Quilio | Cambridge, UK

I build engines, instruments, and languages from scratch.

The Story

I started out doing contract work in games. Systems, procedural graphics, environments. Got really into the tools side of things and naturally ended up in audio, where I became a top seller on Fiverr building plugins and DSP tools for companies around the world.

Eventually I got tired of solving the same problems differently for each client, so I started Quilio to build the SDKs that solve all of them at once. We now ship commercial audio plugins to 100,000+ musicians worldwide.

I studied Neuroscience at Cambridge and Music Technology at Birmingham Conservatoire. I still write papers on the side. I also run and do calisthenics.

By the Numbers

154    public repos
20+    languages shipped in
200K   lines of C++ in the Quilio SDK
500K+  end users across our plugins
20+    commercial products shipped
7+     research papers (1 published, rest in progress)
3      game engines, from scratch
2      programming languages designed

Things I Built From Scratch

Umbra Engine
_{2D game engine. SDL2, Box2D physics, spatial audio, scripting system, an LLM-playable interface, and 51 built-in example games}

Aetherion
_{3D engine. Vulkan rendering, cascaded shadow maps, GLSL shaders. Recently added macOS Metal support}

IslandBoy
_{ECS survival game. PixiJS + bitECS, 28 game systems, 6-species ecosystem sim with predator-prey and weather}

Quilio SDK
_{200K lines of C++. Audio graph with 30+ effects, ML library with neural voice synthesis, hot reload, automated deployment}

Flow
_{A programming language I'm designing. v0.7.0. Graphics API, autodiff, audio DSP library, memory management}

Flux
_{A Turing-complete graph-rewriting language. Pattern matching on typed port graphs. 16K lines, all tests passing}

tinyML
_{ML framework from scratch in C++. CNNs, Transformers, PINNs, GNNs, RL, Bayesian NNs. SIMD-optimised with a WASM playground}

Temple
_{Async HTTP server in Rust. Axum, Tokio, SQLite, crypto, multi-crate workspace}

Research

I write papers about things I find interesting. The first one is published. The rest are in progress and available on request.

Paper	What Happened
Brain-responsive music neurostimulation	Published in the Journal of Neuro Engineering, 2024
FBIR (reactive DAG optimisation)	Got a 29x speedup. In progress
Adaptive Neural Field (time series)	R² = 0.9999. Beat LSTM, Transformers, spiking nets, everything I threw at it. In progress
RCNet (complex-valued neural networks)	Phase rotation + amplitude modulation. Full ablation study. In progress
ECTM (energy-constrained thought)	What happens when you give a neural network a metabolism? In progress
Proof-Limits	Programmatic epsilon-delta proofs using interval arithmetic. In progress