RESEARCH
I like to make databases faster and more efficient! Recently, I have been working on Data Compression and Vector Similarity Search (ANN) within Vector Databases. I am currently working with Prof. Peter Boncz at CWI as part of the Database Architectures group — where I am doing a PhD.
Before, I worked doing research in Social Data Science and Opinion Mining with Prof. Carmen Vaca at ESPOL University (Guayaquil, Ecuador)
RECENT PUBLICATIONS
ALP: Adaptive Lossless floating-Point Compression
[2023] Afroozeh, Azim, Leonardo Kuffo, and Peter Boncz
ALP is an algorithm to compress float and double data; with an insane decompression speed (sub-cycle performance per tuple). ALP can compress better and faster than Chimp128, Patas, ELF and even Zstd. ALP speed stems from its simple algorithm, its vectorized design (operating on 1024 tuples at a time) and the use of SIMD instructions. ALP is currently being used inside DuckDB; for which I contributed to the implementation.
Mining worldwide entrepreneurs psycholinguistic dimensions from Twitter
[2018] L Kuffó, C Vaca, E Izquierdo, JC Bustamante
This study characterized entrepreneurs from different countries based on the emotions of their speech communicated through social media posts. We found strong correlations between the emotions within the entrepreneurs’ posts and various entrepreneurship health indexes within their countries.
Back to# 6D: Predicting Venezuelan states political election results through Twitter
[2017] R Castro, L Kuffó, C Vaca
Back to #6D proposes a framework to make estimations of election results at a country scale by analyzing people’s opinions in a social network. In this study, we tested our framework by estimating the results of the Venezuelan Parliamentary Elections (2016) using solely Twitter posts published before the voting day.