Leonardo Cofone
AI Researcher & Engineer | 17 years old | Deep Learning, Machine Learning, NLP & LLMs
Main links
PROFESSIONAL PROFILE
Through self-driven study, I have developed strong skills in Python programming and Artificial Intelligence. I have built and experimented with Machine Learning, Deep Learning, and NLP models, including computer vision systems, predictive analytics, and large language models. I enjoy building practical AI systems, working with both pre-trained and custom models, and developing intelligent agents capable of learning, reasoning, and solving real-world tasks. I have worked with both structured and unstructured data, applying AI to build real-world applications across different domains.
Show my full CV
WORK EXPERIENCES
-
Mentee – Lead The Future Mentorship
Among the few Italian students selected to be mentees for LeadTheFuture, a leading mentorship non-profit organization for students in STEM, with acceptance rate below 20%. LeadTheFuture empowers top-performing students to achieve their goals and contribute to their communities by giving them one-on-one guidance from high-impact mentors coming from the world's leading STEM innovation hubs such as Silicon Valley and CERN. -
Guest Researcher – Max Planck Institute for Intelligent Systems
Research collaboration in the Deep Models and Optimization group at MPI-IS, working on looped models and the efficiency of deep architectures.
I worked through recent literature on efficient deep architectures and presented my own paper and independent work to the group, discussing directions for improving how deep models scale. Working in that environment and getting feedback from researchers with far more experience than me changed how I approach research problems. -
Python Developer – Airpim Srl
Designed and built a zero-dependency Python tool for automatic memory leak diagnostics.
The hard part was telling real leaks apart from normal allocation noise without fixed thresholds, so the tool adapts its statistics to the process it is watching. When it flags something, it walks the object graph to find what is actually holding the reference, instead of just reporting that memory is growing.
No external dependencies, so it drops into any codebase. Validated in production. -
AI Developer – EgoLog Srl
Built and deployed AI agent systems in n8n for social media management, data enrichment, and lead generation.
The work was end to end: understanding what the client actually needed automated, designing the agent workflows, integrating them with the systems already in place, and keeping them running once they were live. -
AI Developer – Top Evolutions Srl
Worked on AI automation for businesses in the payments space, building prompt-driven systems that matched merchants with the payment provider that actually fit them.
Alongside that, automation work on lead generation. Early on I also trained models on the client's data, before the work shifted toward prompt engineering: getting the model to reason over real business constraints instead of producing plausible-sounding recommendations. - Developed multiple personal AI projects, including deep learning models for computer vision, predictive analytics, and natural language processing, demonstrating hands-on experience across both research and applied AI.
EDUCATION AND LEARNING
- Currently studying at Galileo Galilei Scientific High School, Trento – Expected graduation: June 2028
- Actively expanding my knowledge in Artificial Intelligence through hands-on projects, online specializations, advanced courses, competitions, and AI-related books.
- See below for a detailed list of completed certifications and courses.
COMPUTER SKILLS
Programming: Python (advanced), C++ (basic). Web: HTML, CSS, JavaScript (basic).
Artificial Intelligence & Machine Learning: Hands-on experience with supervised and unsupervised learning, deep learning, computer vision, predictive modeling, and natural language processing. Experience in building AI systems and LLM-based applications using both pre-trained and custom models.
Relevant Projects & Applications: Developed machine learning pipelines for classification, regression, clustering, and data analysis tasks. Built neural network models for computer vision and NLP applications, and experimented with LLM-based systems and AI agents using prompt engineering. Applied these techniques in personal projects, competitions, and real-world datasets, focusing on model evaluation and performance improvement.
CERTIFICATIONS
Certifications in Machine Learning, Deep Learning, Computer Vision, and NLP, focused on both theoretical foundations and hands-on implementation using real-world datasets.
Machine Learning Specialization
Andrew Ng - Stanford University, 94h
Foundations of supervised and unsupervised learning with practical implementation in Python.
Supervised Machine Learning
Andrew Ng - DeepLearning.AI, 33h
Core concepts in regression, classification, and evaluation metrics.
Unsupervised ML, RL, Recommenders
Andrew Ng - DeepLearning.AI, 27h
Clustering, recommender systems, and reinforcement learning fundamentals.
Advanced Learning Algorithms
Andrew Ng - DeepLearning.AI, 34h
Neural networks, optimization, and regularization techniques.
Computer Vision & Image Processing
IBM, 22h
Image processing, detection, and classification using OpenCV.
Machine Learning Capstone
IBM, 20h
End-to-end machine learning pipeline on real-world datasets.
Machine Learning with Python
IBM, 20h
Model development using Python and scikit-learn.
PROJECTS
Here are some of my recent AI projects. More work available on my GitHub and Kaggle profiles.
DSALT
Research Project · Transformer ArchitecturesResearch on sparse attention mechanisms for improving long-context Transformer efficiency and representation quality.
See more
DSALT: Dynamic Sparse Attention with Landmark Tokens
DSALT is a research project on Transformer attention mechanisms that investigates the structural limitations of dense self-attention in long-context settings. In particular, it analyzes how uniform softmax weighting can lead to cumulative representational noise and progressive homogenization of token embeddings across layers.
To address this, DSALT introduces a dynamic sparse attention mechanism that combines local windowed attention with globally selected landmark tokens. Landmark selection is performed using a hybrid energy-based scoring function that balances representational magnitude and output relevance.
The objective is to reduce redundant token interactions while preserving long-range dependencies, improving both representational quality and computational efficiency. The project also explores the relationship between attention structure, rank dynamics, and information flow in deep neural networks.
LerriAI
AI Assistant · LLM · ProductivityA powerful personal AI assistant to help you organize your life, manage everything with ease. (Currently offline)
See more
LerriAI is built around a simple idea: most people don’t lack ability, they struggle with organization. Life throws countless tasks, deadlines, emails, routines, goals, and unexpected problems at us, and even the most motivated individuals can easily feel overwhelmed. LerriAI exists to manage that complexity. It is not just a chatbot or a simple assistant, it is a living system of intelligent agents that work together to structure your day, support your decisions, and keep your personal world running smoothly. Each agent has a specific role, analyzing your habits, priorities, and preferences, planning, reminding, optimizing, summarizing, and tracking progress, while communicating through a shared intelligence layer. In this way, LerriAI turns chaos into order, complexity into clarity, and everyday distractions into focused, actionable guidance, helping you stay on track with what truly matters.