Work

Three projects, each started because I wanted to understand something properly.

Stivium

A lightweight desktop IDE for Verilog and RTL, with an AI assistant built in.

Rust · Tauri · React · TypeScript · Monaco · Icarus Verilog

Stivium is a lightweight desktop IDE I built for Verilog and RTL development. It brings the editor, simulator, terminal and an AI coding assistant together in a single window, so a module can be written, simulated and debugged without ever switching tools. It is built with Rust, Tauri and React: one click compiles the live editor buffer and simulates it with Icarus Verilog, and compiler errors and simulation output stream straight into an integrated console.

The AI panel is multi-provider. It supports Anthropic, OpenAI-compatible APIs and local models through Ollama, and API keys stay safely in the OS keychain. Around it sit a Monaco-based editor, a real pseudo-terminal and extensible language server support, which together make it feel like a proper development environment rather than a toy.

I finished it with a custom frameless interface, an animated boot screen and a small desktop companion for code review. Tools you use every day should have a bit of personality.

Neural Engine

An independent archive of neuromorphic and brain-inspired computing research.

with Tejas Kapoor

Neural Engine is an independent archive of more than 110 research works across ten domains of neuromorphic and brain-inspired computing, built together with Tejas Kapoor. It puts the field’s papers in one organised place, instead of a hundred open tabs.

All of the research in it belongs to its original authors. The archive’s job is simply to help you find their work.

All-Digital Resonate-and-Fire Neuron

A neuron that only fires at the right rhythm, rebuilt from digital logic.

in progress with Tejas Kapoor

A resonate-and-fire neuron, a model proposed by Eugene Izhikevich, does more than add up its inputs until a threshold is crossed. Its membrane potential oscillates, so it reacts most strongly to input that arrives at its own preferred rhythm and shrugs off the rest. It behaves less like a switch and more like a tuning fork.

I am implementing the all-digital version designed by Trung-Khanh Le, Trong-Tu Bui and Duc-Hung Le. Their paper turns the model’s equations into digital pulses and builds the neuron without floating-point operations or multipliers, using just 23 flip-flops on an FPGA. It is my reference for the hardware, with Tejas Kapoor alongside me for the build.

Alongside the implementation, I adapted the paper into an interactive walkthrough for the web, so the idea can be explored before a single line of hardware description.

Based on Modeling and Designing of an All-Digital Resonate-and-Fire Neuron Circuit by Trung-Khanh Le, Trong-Tu Bui, Duc-Hung Le (IEEE Access). All credit for the design belongs to them.