Building innovative solutions for complex visual problems, from object detection pipelines to production cloud and edge deployment. Currently Lead CV Engineer at Technology Data Reply IT, Milan. DM me to grab a coffee! ☕ 😊
Proficient in:
Federico Favia
Real-time hand segmentation using deep learning — my master's thesis project, developed during an industry research internship at ManoMotion AB, as part of a double master's degree with EIT Digital, jointly awarded by KTH Royal Institute of Technology and the University of Trento. Several encoder-decoder segmentation architectures (U-Net, ICNet, RefineNet, DeepLabV3+, PSPNet) are implemented, trained and benchmarked, trading off accuracy, inference speed and model size for real-time mobile AR use.
View on GitHub ↗
End-to-end ML project predicting product category from name + brand across languages (TF-IDF/Logistic Regression + fine-tuned multilingual DistilBERT approach), served via FastAPI with Docker.
Explores image segmentation through both classical image-processing techniques and modern deep learning approaches — from thresholding, watershed and contour-based methods to neural network segmentation models, applied to real-world imagery.
Bachelor Thesis at Xtensa Srl — comparing deep learning based tracking with YOLOv3 against classic algorithms like Mean-shift for football player detection and tracking.
Estimates naval traffic in Venice from a webcam video of the Grand Canal using motion detection algorithms — adaptive background subtraction, morphological filters, and blob detection to count boats and compute flow rate, integral average, and median.
Image classification via a Visual Search System built on a hierarchical K-means tree. Features represented by SIFT keypoint descriptors, scored with TF-IDF using the Bag of Visual Words model.
A two-part computer vision project combining face detection, face matching and facial attribute analysis (age, gender, race, emotion) with intrusion detection for CCTV-style video surveillance scenarios.
End-to-end custom object detection pipeline built with Ultralytics YOLOv8 — dataset preparation via Roboflow, fine-tuning, evaluation, and inference on images and video, including object tracking.
Button-triggered safety check-in system on Arduino: press once to send a check-in SMS, press again if unanswered to broadcast the last known GPS position and a Google Maps link to emergency contacts.
Lightweight Python tool that converts structured Excel task list into a clean, publication-ready Gantt chart for PMs and engineers without access to Jira.
Investigation of two popular scale-invariant feature detectors — SIFT and SURF — for feature extraction and image matching using nearest-neighbour distance ratio.
Interactive 3D Solar System demo with keyboard navigation and .jpg textures on planets, rendered with GLEW and GLUT.
Multimedia Data Security project detecting and localizing tampered image artifacts using PRNU-based and JPEG-based forensic techniques.
Personally created and designed the logo for the Engel startup during the Startup Lab Course at Clab — University of Trento, using Adobe Illustrator.
Let's build something
worth seeing.