Loris Bazzani
Adjunct Professor at University of Verona
Berlin, Germany
Summary
👤
Senior🎓
Top SchoolLoris Bazzani is an AI research leader and adjunct professor with 12+ years of hands-on experience building and shipping multimodal and vision-language systems at scale. He drove foundational deep learning efforts at Amazon—pioneering video understanding, large multimodal models, and production features like live sports highlights and virtual try-on that reached millions and generated nine-figure impact. Equally at home in academia and industry, he has 50+ publications and patents, teaches data visualization, and collaborates across international research networks including ELLIS. His engineering work extends into open-source contributions for sequence-to-sequence frameworks (e.g., image captioning in awslabs/sockeye), reflecting a practical focus on integrating vision and language pipelines. Loris excels at turning cutting-edge research into product-relevant systems by leading cross-functional teams and scaling science organizations. Outside work he cultivates creativity and resilience through music, running, and meditation, feeding a multidisciplinary approach to AI.
12 years of coding experience
2 years of employment as a software developer
PhD, Computer Vision and Machine Learning, PhD, Computer Vision and Machine Learning at Università degli Studi di Verona
Italian, English, German, venetian
Github Skills (15)
model-building10
sequence-to-sequence10
pytorch10
machine-learning10
computer-vision10
image-text10
deep-learning10
image-annotation10
neural-machine-translation10
python10
modeling10
model-driven10
model-driven-development10
inference9
data-loading9
Programming languages (1)
Github contributions (5)
awslabs/sockeye
Feb 2018 - Jul 2019
Sequence-to-sequence framework with a focus on Neural Machine Translation based on PyTorch
Role in this project:
ML Engineer
ML EngineerContributions:7 commits, 9 PRs, 97 comments in 1 year 5 months
Contributions summary:Loris primarily focused on developing and implementing features for image captioning within the sequence-to-sequence framework. Their contributions included creating new modules for image feature extraction, data loading, training, and inference pipelines. The user also introduced zero-padding capabilities for image features, added scoring functionality, and made versioning updates. This work demonstrates a focus on integrating computer vision and natural language processing components within the existing NMT framework.
sequencetranslationsockeyeneural-machine-translationseq2seq
lorisbaz/self-taught_localization
Jan 2014 - Mar 2015
Contributions:70 commits, 4 pushes in 1 year 1 month
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.