projects I've worked on, ranging from research code to side experiments.



scaling recipes for bio foundation models
aug 2026bioinformatics, machine learning, scaling laws
investigated how deep learning models scale on transcriptomics data, quantifying power-law relationships between model size, compute, data and pretraining loss, downstream performance.



integrated research environment (ire)
jun 2026desktop app, rust, ai agents
a local-first desktop environment for researchers that keeps code, literature, experiments, and ai collaboration seamlessly connected, featuring a persistent git-tracked wiki, non-blocking experiment tracking, and a native mcp bridge for agentic ide integration.


autoresearch hft
may 2026quantitative finance, machine learning, ai agents
self-looping agents tested 30 high-frequency trading strategies in under 48 hours, going from losing money to earning over €25k/month on nanosecond precision order book data.

multi-modal real estate search
apr 2026python, mcp, hackathon
built a multi-stage, multi-modal search pipeline that transforms natural language queries into ranked real estate listings for the swiss market, deployed as an mcp server on aws enabling any compatible llm to query real-time listings.

diffusion language models are natively length-aware
mar 2026natural language processing, language modelling
showed that diffusion language models encode target output length and used it to make inference significantly more efficient.

knowledge graph of the italian law
aug 2025python, neo4j, knowledge graph
designed and developed a proprietary etl pipeline to construct a knowledge graph from over 5,000 unstructured italian supreme court's rulings, and implemented an out-of-core, efficient pagerank-based algorithm to assign relevance scores to graph nodes.

do ai and human experts see pneumonia the same way?
may 2025computer vision, explainable ai, medical imaging
compared cnns with experienced professionals on pediatric chest x-rays to test whether deep learning models identify pneumonia using the same visual cues as doctors, finding that they do not and that high accuracy doesn't guarantee human-interpretable reasoning.

wordle solver
feb 2025natural language processing
optimal strategy solver for wordle that achieves 100% success rate with an average of 2.89 guesses in 0.5 seconds.


bsc thesis
jul 2024bioinformatics, machine learning
enhanced scgpt transformer model for gene expression prediction by incorporating prior biological knowledge of transcriptional regulation.


daily spin bot
dec 2023python, browser automation
automated browser bot for maximizing rewards in snai daily spin game, available as executable.