AI Engineer | Applied ML | Cloud Systems
Based in Turin, Italy
Hello! I am a passionate data scientist focused on extracting meaningful insights from complex data to drive informed decisions. I build practical ML systems with a strong focus on Large Language Model deployment and optimization.
I enjoy turning complex problems into simple, intuitive experiences, and I care about taking ideas all the way to production-ready products.
Built an AI-powered desktop video translation application that automates speech-to-text, subtitle generation, subtitle editing, and final video export with hardcoded subtitles. The project combined Electron for the user interface, Python for the processing pipeline, AWS Transcribe for transcription, AWS Translate for multilingual subtitle translation, and FFmpeg for audio extraction and video rendering. I also developed a timeline-based subtitle editor, custom font support, and a smoother workflow for preparing translated video content.
View Project
This is a full-stack personal finance tracker built with Django, Ionic React, TypeScript, and Capacitor. The app allows users to securely log in, track income and expenses, view transaction history, edit or delete entries, and monitor financial summaries such as totals, averages, and counts. I worked on both the backend API and the mobile-friendly frontend, connecting authentication, CRUD workflows, and data handling into a clean cross-platform experience.
View Project
Built an AI-powered mental health triage and booking system using Django and a 3-agent workflow: adaptive intake, specialist matching, and automated scheduling. The platform analyzes symptoms with safety guardrails, matches patients to the best specialist based on expertise and location, then books the most suitable time slot based on urgency and preference. It includes dynamic form behavior, session-based flow orchestration, JSON/data pipeline handling, and confirmation email integration through AWS Lambda. (Reply Agentic Challenge)
View Project
We tested multiple models, including both proprietary (GPT, Claude, Gemini) and open-source (Code Llama, SynthiaIA, Toppy) LLMs. Our pipeline took a novel two-stage approach: instead of asking the model to generate both logic and syntax, we separated concerns. The LLM would first generate the logical structure of the ER model, and then a separate syntax module would convert that into a Designer.io-compatible JSON format.
View Project
Designed a computer vision system that could identify different types of waste in images and run efficiently on devices with limited computational power. To solve this, we focused on semantic segmentation, assigning a class label to every pixel in an image for smart waste sorting in real-time and edge-device deployment. We evaluated three resource-efficient models tailored for speed and deployment feasibility.
View Project
Converted audio signals into Mel spectrograms, then split them into blocks to extract summary statistics. After filtering inaudible samples and trimming silent sections, we transformed spectrograms to a logarithmic scale aligned with human sound perception. We used a classical machine learning pipeline with hand-engineered features and models such as Random Forest and SVM, where Random Forest performed best after grid-search tuning.
View Project
Avatar assistant in progress