Models that are measured, not just demoed.
We build predictive models, computer vision systems and LLM-powered tools. We benchmark them honestly against simpler alternatives, so you know what you're getting.
Discuss your dataWhat we do
Predictive models
Regression and classification on tabular data: forecasting, scoring, demand and risk estimation, with gradient boosting and neural networks.
Ensemble methods
Research-level work in dynamic ensemble selection, choosing the best model for each prediction. In our benchmarks across 35 datasets, up to 27% lower error than the single best model.
Computer vision
Image classification and analysis with custom convolutional networks and CLIP, reaching over 90% accuracy on plant classification in our own project work.
LLM and RAG systems
Retrieval-augmented generation: assistants that answer from your documents, with vector search and cloud deployment.
Data preparation
Collection, cleaning, augmentation and feature engineering. This is the part that decides whether a model works.
Tuning and benchmarking
Systematic hyperparameter optimisation and fair comparison against baselines, so improvements are real.
Tools and libraries
Our core stack is Python and C++.
AI and security together
If you're adding a chatbot or LLM feature to your product, we can build it and test it for prompt injection and data leakage. See Security Testing.
Credentials
National competition results
Bronze medal at the Italian Olympiad in Informatics (top 0.3% of ~15,000) and 8th place nationally at the Italian Olympiad in Artificial Intelligence.
Open-source research
Author of an open-source dynamic ensemble selection library, including novel methods for regression developed with an active published researcher.
Have data and a question?
Describe the problem and the data you have. We'll tell you honestly whether machine learning is the right tool.
info@xano.it