About
Staff Research Engineer at Qualcomm AI Research (November 2019 - Present)
I design and implement novel Deep Learning methods across multiple domains including Perception (visual and RF), Model Efficiency, and Agentic AI (LLMs). My work has led to significant innovations:
- Invented a personalized biometric neural network classifier that improved accuracy from 90% to 99% AUC
- Implemented Neural Kalman Filter and Weighted Least Squares methods for GNSS localization, reducing positioning error by 50%
- Designed a Dynamic Pruning algorithm for recent LLMs (Phi 3, Llama 3), improving throughput by 40% and memory footprint by 45%
- Introduced structured representations in AI Agents, reducing model footprint by 30x
- Integrated solutions in products released to customers (Samsung, Google) and internal departments
- Filed 11 patent applications (1 granted, 10 pending)
Research Interests
Large Language Models, Model Efficiency, Perception, Agentic AI, GNSS Localization, Graph Neural Networks
Skills & Technologies
Python (PyTorch, Hugging Face, vLLM, NumPy, Matplotlib, Pandas, Hydra), Bash, Git, Docker, Run:ai, MATLAB, C++, Java
Education
MSc in Artificial Intelligence
University of Amsterdam
September 2017 - September 2019
Graduated Cum Laude, GPA 8.8/10 (A+)
Graduate Teaching Assistant (October 2018 - June 2019): Teaching Assistant for core courses Machine Learning 1, Deep Learning, and Information Retrieval. Held lab sessions, prepared and corrected exams, homework, and lab assignments.
Main Courses: Machine Learning, Computer Vision, Natural Language Processing, Deep Learning, Reinforcement Learning, Information Retrieval
BSc in Computer Science
University of Trento
September 2014 - July 2017
Graduated Cum Laude, GPA 110/110 (A+)
Main Courses: Algorithms and Data Structures, Programming, Calculus, Linear Algebra, Probability and Statistics, Logic, Databases