Website: https://marcrulo.github.io/MonthlyProjects/
MonthlyProjects is a lightweight archive of monthly blog posts, experiments, and project write-ups.
Building a fully-connected neural network in pure NumPy to test my skills, for when I can't simply ask Google or AI for help at the slightest inconvenience.
🎮 Game RL
A 2D bouldering environment in PyGame/Gymnasium and a PPO agent that learns to climb via tuned rewards and movement constraints.
An LLM-based bot for our Discord server with humor derived from content from our childhood.
Hands-on tests of Meta's SAM Audio for stem separation, with local setup notes and cherry-picked instrument demos.
🍽️ Food Pairing
Exploring FlavorGraph and ingredient embeddings to propose pairings, plus lessons from graph vs. vector similarity.
Building the preference model, uncertainty sampling loop, and ranking pipeline, then shipping weekly email recommendations.
From scraping fixes to a swipeable web app, plus tradeoffs between AI app builders and a Flask-based prototype.
Scraping weekly grocery flyers, enriching item metadata, and generating playful bios to bootstrap preference data.
🌐 LiDAR
Using an RPLIDAR to detect a circular boundary and hole width with DBSCAN and circle fitting, plus rigging lessons.
Building a Chrome extension that visualizes prerequisite graphs by scraping DTU course pages and layering the graph.
Competition write-up covering the three challenges, especially the RAG-based medical fact-checking pipeline.
Cloud-hosting the Danish court-case RAG system on Azure, with costs, endpoint setup, and evaluation examples.
Collecting court documents, chunking and embedding Danish text, and building a FAISS-backed retrieval pipeline.
Using diffusion with classifier guidance to remove spurious visual shortcuts and improve lesion classification.
Training a custom embedding for Charizard in Stable Diffusion v1.5, with data curation and tuning notes.
🌐 Website
Setting up a Jekyll blog on GitHub Pages, customizing the theme, and documenting the monthly-project workflow.