Make your own tools.

FEAx is a Python framework for procedural Finite Element Analysis (FEA). Built on top of compas_fea2, the open-source library I developed during my PhD at ETH Zurich, it represents its natural evolution. FEAx was designed to provide a public API for FEA that is backend-agnostic. The result is a plug-in architecture in which meshing, analysis, data storage, and visualization can be fully interchanged to suit the user’s preferences.

reentry is a CLI tool for multi-repo workspace orchestration that helps you context-switch back into complex projects. Builds a cross-repo dependency graph from Python AST analysis with blast-radius computation, indexes your workspace into ChromaDB vector embeddings for semantic code search via natural language queries, classifies commits and detects repo state anomalies (dirty trees, upstream drift, stale sessions). Everything is exposed through an MCP server so AI assistants can reason about your workspace and rank repos by urgency — so you can run reentry resume and get a full re-entry brief of where you left off, what changed, and what to do next.

Taltal is a universal, framework-agnostic evaluation engine for reproducible, auditable assessments of any entity (companies, projects, systems) against versioned scoring frameworks. I built it to guide VAULTED’s strategy.

A lightweight, self-hosted VDR built to manage due diligence without the cost and complexity of enterprise platforms. Designed for early-stage deep-tech companies, where tight information control and investor access matter. I built it in one day after being quoted $5,000 for what was essentially a glorified Dropbox.

The financial modelling library built from scratch for VAULTED. Handles unit economics, P&L forecasting, production scaling scenarios, and cash flow projections — structured around the specific dynamics of a hardware-software deep-tech company.

An early collection of Python scripts to automate repetitive tasks across engineering workflows. Largely superseded by AI-native tools, but the instinct behind it — scripting away friction — is still the same.

Built with the help of ChatGPT and Claude, I thought myself next.js and typescript and had a lot of fun building it. It combines my passion for design and technology. I hope you like it too