AI Engineer specialized in Computer Vision and Embedded AI, from R&D to deployment on embedded targets.

I design real-time vision systems under tight power and compute constraints (quantization, TensorFlow Lite, ONNX, NPU acceleration), and I explore AI agents, multimodal systems and privacy-first edge AI. Dual Master's degrees in AI & Visual Computing.

Mission: build robust, efficient and impactful AI that runs where it matters, from cloud to edge.

🔹 UniCorrn, official codebase of the CVPR 2026 paper "Unified Correspondence Transformer Across 2D and 3D" (Northeastern University Visual Intelligence Lab, with Adobe Research).

Contributed PR #5 (merged): lazy loading of optional 3D dependencies and a pure-PyTorch kNN fallback, making the codebase portable to Windows.

PyTorch · 3D Vision · Correspondence Matching · Open Source

🔹 Smart Vehicle Detector Real-time vehicle detection pipeline combining YOLO-based detection with a Vision-Language Model for fine-grained classification and structured JSON outputs.
YOLO · VLM · Real-time CV · Docker

🔹 UNETR 2D Medical Segmentation From-scratch PyTorch implementation of a 2D variant of the UNETR architecture, with reproducible training and tooling for biomedical datasets.
PyTorch · Transformers · Segmentation · Medical AI

🔹 EasyNodule Desktop application for AI-driven pulmonary nodule classification from segmented CT scans, built on a custom CapsNet model with patient management and PDF reports.
CapsNet · TensorFlow · Desktop App · Medical AI

🔹 HealthRAG Medical chatbot powered by Retrieval-Augmented Generation: PDF ingestion, FAISS vector store, LangChain and a hosted LLM.
RAG · LangChain · FAISS · LLM

Core & Vision

Edge & Embedded

LLM & Agents

MLOps & Data

"AI isn't just about intelligence: it's about creating systems that serve and empower people."