Skip to content
View apiantonio's full-sized avatar
🎯
Focusing
🎯
Focusing
  • Università degli studi di Salerno
  • 01:45 (UTC +02:00)

Block or report apiantonio

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
apiantonio/README.md
Typing SVG



LinkedIn Email GitHub followers


👨‍💻 About Me

I'm an AI Engineer and MSc student in Computer Engineering at the University of Salerno, specializing in Artificial Intelligence. I design and build end-to-end intelligent systems — from production-grade Agentic RAG pipelines and LLM applications to state-of-the-art Computer Vision and Generative AI architectures.

My software engineering background means I don't just prototype — I build systems that are modular, maintainable, and ready for real-world deployment.

  • 🔭  Focus areas: Agentic LLM Systems · Generative AI · Real-time Computer Vision
  • 🎓  MSc in Computer Engineering (AI) @ University of Salerno — graduating 2025/26
  • 🌍  Based in Campania, Italy · Open to relocation / remote
  • 📬  Reach me: apiantonio.aa@gmail.com

🚀 Featured Projects

Production-grade LLM system for the DIEM Department at the University of Salerno.

An intelligent assistant that autonomously queries multiple vector store collections, rewrites ambiguous queries, and generates grounded multilingual answers — with full guardrail and RAGAS evaluation pipelines.

Highlights:

  • Agentic parallel tool-calling (LangChain) — LLM autonomously selects knowledge collections
  • Multi-threaded BFS web crawler feeding 3 ChromaDB collections
  • Smart conversational memory with semantic filtering
  • 3-stage Llama 3.3 70B guardrail system (injection · toxicity · hallucination)
  • Full RAGAS evaluation pipeline → JSON/CSV/Excel reports

Python LangChain ChromaDB Ollama Groq Streamlit

Computer Vision pipeline for the SoccerNet Benchmark — A.Y. 2025/26 @ UniSa.

Full Tracking-by-Detection system combining high-recall detection with motion-based tracking and semantic field masking.

Results on Official Test Set:

Metric Score
🏆 HOTA 0.742
DetA 0.863
AssA 0.636
PTBS 1.527
  • YOLOv11x at 1088px with adaptive HSV/LAB field masking
  • BoT-SORT with Sparse Optical Flow (GMC) — no visual Re-ID needed
  • Modular YAML-configurable pipeline (tracking · behaviour · eval · viz)

Python YOLO OpenCV PyTorch

Generative AI research comparing three architectures for conditional face synthesis on CelebA.

A comparative framework for controlled face generation (gender · expression · age) benchmarking three distinct generative paradigms.

Architectures implemented:

  • Vision Mamba (SSM) — custom autoregressive implementation with parallel scan (pscan)
  • Conditional DDPM — U-Net with sinusoidal time encoding + Classifier-Free Guidance
  • Conditional VAE — symmetric encoder-decoder with residual skip connections

Python PyTorch HuggingFace CUDA

Semantic segmentation optimized for resource-constrained on-board deployment.

Semantic segmentation of rural aerial imagery using a ResNet18 encoder coupled with a U-Net decoder architecture, designed for efficient inference on embedded hardware.

Key aspects:

  • ResNet18 pretrained backbone with U-Net skip-connection decoder
  • Optimized for on-board inference (low memory footprint)
  • End-to-end pipeline in PyTorch with Jupyter-based experimentation

Python PyTorch Jupyter


🛠️ Core Tech Stack

🤖 AI · LLM · Machine Learning

Python PyTorch HuggingFace LangChain scikit-learn Pandas NumPy Jupyter

👁️ Computer Vision

OpenCV Ultralytics CUDA

🗄️ Vector Stores & LLM Infra

ChromaDB Ollama Groq Streamlit

⚙️ Engineering & DevOps

Git Docker Linux C


📊 GitHub Stats

Antonio's GitHub Metrics

📫 Let's Connect

I'm actively looking for AI Engineer / ML Engineer opportunities (full-time or stage). If you're working on something interesting in the LLM, GenAI, or Computer Vision space — let's talk.

LinkedIn Email

Pinned Loading

  1. LLM_powered_chatbot_DIEM LLM_powered_chatbot_DIEM Public

    An Agentic RAG-powered chatbot delivering intelligent, fact-grounded and context-aware answers for the DIEM Department at the University of Salerno.

    Python 4

  2. SoccerNet-tracking_AV2025-26 SoccerNet-tracking_AV2025-26 Public

    Robust Player Tracking & Behavior Analysis pipeline for SoccerNet Benchmark. Features YOLO11x, BoT-SORT (GMC), and Adaptive Field Masking. Artificial Vision Project 2025/26 @ Unisa.

    Python 4

  3. Attribute-Guided-Face-Generation-A-Mamba-Diffusion-and-VAE-Framework Attribute-Guided-Face-Generation-A-Mamba-Diffusion-and-VAE-Framework Public

    Forked from SimoneFaraulo/Attribute-Guided-Face-Generation-A-Mamba-Diffusion-and-VAE-Framework

    A comparative Generative AI framework for conditional face synthesis. Features a custom autoregressive Vision Mamba (SSM) architecture, benchmarked alongside Conditional DDPM and VAE models on the …

    Python 1

  4. GenerativeAI-Foundations GenerativeAI-Foundations Public

    Forked from SimoneFaraulo/GenerativeAI-Foundations

    This repository contains a collection of PyTorch scripts that implement and explore the fundamental concepts of probability, information theory, and generative modeling from scratch. It serves as a…

    Python 1

  5. Rural-Images-Segmentation-on-board Rural-Images-Segmentation-on-board Public

    Jupyter Notebook 1