Projects

IcyMaps generates detailed 2D worlds from compact inputs using procedural terrain and biome systems. The project is evolving toward a simulation environment where organisms can interact with their surroundings, reproduce, adapt, and die across accelerated timelines.

• Rust
• Procedural Generation
• Simulation

IcyGut explores representation learning across metagenomic, metabolomic, and clinical data using modality-specific neural encoders and learned feature fusion. The project focuses on disease-state classification, patient-level latent trajectories, missing-modality training, grouped validation, representation analysis, and model explainability.

• Python
• PyTorch
• Machine Learning
• Multi-omics
• Representation Learning

At DFKI's CuraHack, we worked with 16S gut microbiome sequencing data to predict a health-risk outcome. The project emphasized robust validation and explainability, showing how biological variability and high-dimensional microbial data can make careful modelling decisions as important as raw predictive performance.

• Python
• Scikit-learn
• PyTorch
• 16S Sequencing

Icy Editor is a React and TypeScript application for building structured question papers through manual or automatic question selection. It combines reusable question banks, rich-text editing, configurable exam structures, live previews, saved-paper management, and DOCX export.

• React
• TypeScript
• Education
• DOCX

IcyMed enables clinicians to create reusable clinical forms, publish secure one-time patient intake links, and review submitted data from protected workspaces. The MVP includes clinician authentication, configurable question types, patient consent capture, form templates, and Supabase-backed storage with row-level security.

• React
• TypeScript
• TanStack
• Supabase
• Healthcare