Hybrid RAG for corporate documentation
An auditable, end-to-end RAG system: it processes heterogeneous documents, combines BM25 with vector search, reranks evidence and preserves page-level traceability so every answer can be verified.
Main projects / 05
Projects that explain how I work.
I selected the work that best represents my interests and abilities. Each case explains the problem, how it works, the technologies I used and my responsibility.
Read project
Airfryer: computer vision for intelligent cooking
An ambitious R&D project connecting visual perception with cooking planning: it analyses a basket, separates individual pieces and proposes cooking times and temperatures. It combines deep learning, computational geometry and interpretable rules in an architecture aimed at universal automatic cooking.
TwinPhoto
A desktop product combining perceptual hashing, an incremental SQLite index, parallel processing and a native Rust bridge to review and safely remove duplicates—even from USB-connected phones.
Bitcoin sentiment analysis
An experimental pipeline for an especially noisy signal: it turns text into semantic embeddings, combines two LightGBM ensembles and translates each classification into an explanation with visible confidence, signals and limitations.
Professional interview assistant
A generative AI assistant designed as a public service rather than an isolated demo: controlled retrieval, NDJSON streaming, sessions, queueing, usage limits and privacy sanitisation in one conversational experience.