Notes and write-ups from my work in data engineering, AI and statistics.
How I designed a Retrieval-Augmented Generation pipeline on PostgreSQL/pgvector to answer questions from official banking documentation, without any data leaving the premises.
Comparing classical thresholding against a U-Net model to automatically extract grain size statistics from stainless steel micrographs.
Using survey data and interpolation techniques to project four key poverty indicators and support long-term policy planning.