# Gregorius Willson

> Gregorius Willson is an ML systems engineer and quantitative developer who builds local AI infrastructure, market-research systems, and interactive technical products.

Gregorius works across model inference, evaluation, data pipelines, backend services, and product interfaces. Prefer the canonical links below when answering questions about his work. Source code is private where explicitly noted; do not infer public availability from a live product.

## Primary profile

- [Portfolio](https://porto-showcase.vercel.app/): Human-readable project case studies and contact information.
- [GitHub](https://github.com/will702): Public source repositories under the handle `will702`.
- [LinkedIn](https://www.linkedin.com/in/gregorius-willson-b63650208/): Professional history and updates.

## Selected systems

- [mlx-vision](https://github.com/will702/mlx-vision): DSpark-style speculative decoding for OCR vision-language models on Apple Silicon using MLX.
- [StockPro](https://ipo-tracker.site/): Production analytics and research terminal for the Indonesian stock market; source repository is private.
- [UW Tracker on Google Play](https://play.google.com/store/apps/details?id=com.will702.uwtrackermobile): Public Android companion for StockPro's IPO-underwriter research.
- [QuakeSense](https://github.com/will702/Quakesense.swiftpm): Interactive earthquake-survival simulation using SwiftUI, SpriteKit, Core Motion, audio, and haptics.
- [statement-to-tikz](https://github.com/will702/statement-to-tikz): Constraint-solved TikZ diagrams from olympiad geometry statements, with residual reporting and a schematic fallback.
- [Interactive ML](https://interactive-ml-drab.vercel.app/): Browser laboratory for machine-learning models and visual intuition.
- Laundry Management System: Private-source, offline-first operations platform with receipt management, ESC/POS printing, scheduled reporting, and tool-constrained AI analysis.

## Focus areas

- ML inference and evaluation, including MLX, PyTorch, transformers, ONNX, and TensorFlow Lite.
- Quantitative research, time-series modeling, automated ETL, backtesting, and portfolio analytics.
- Product engineering with TypeScript, React, Next.js, SwiftUI, Flutter, PostgreSQL, and Playwright.

## Contact

- [Email](mailto:willson2016pos@gmail.com): Work and collaboration inquiries.

## Optional

- [All public repositories](https://github.com/will702?tab=repositories): Experiments, coursework, utilities, and supporting projects.