Ronald Suwandi

Built a rapid prototype integrating Stockfish with LLM-generated explanations to analyze chess moves, aiming to bridge the gap between engine calculations and human reasoning. Found key AI limitations—LLMs often misinterpret board states and struggle with structured logic, leading to subtle hallucinations and explanations. Recently refined by adding FEN for better context and testing a game review feature to improve output quality. Tested GPT-4o and GPT-4o-mini, with GPT-4o performing better but still unreliable. Completed as a proof-of-concept with no plans for further development

Developed a terminal-based chess engine from scratch in Rust using bitboard data structures for efficient move validation. Implemented comprehensive chess rules including complex edge cases for pins, checks, stalemate detection, and en passant captures. Built a terminal UI using ratatui with custom PNG rendering, focusing on PGN notation input to improve tournament preparation skills

Developed a React Native app for iOS that allows users to quickly log their weight into the Health App, featuring a simple and intuitive design.

Built a Clojure-based neural networks library on top of the core.matrix package. Inspired by Andrew Ng's Machine Learning course, the library is modular, allowing for easy swapping of optimizers, activation functions, stopping conditions, and cost functions.

Created a web application that helps users decide what to wear based on the current weather. Developed an optimizer called MACHO, based on Ant Colony Optimization algorithms, to quickly calculate the best clothing combinations. Featured in both StartupsFM and StartupDaily.

A simple mobile-first webapp that translate the sentence into a hashtag format and also gives you some suggestion keywords. Powered by jQuery and Big Huge Thesaurus.