Alessio Pepe's CV

• LinkedIn: pepealessioo
• GitHub: pepealessio

Summary

AI Research Engineer specializing in generative and agentic AI, adversarial robustness, and edge AI. Proven track record of turning research prototypes into deployable systems for defence and security.

Experience

Artificial Intelligence Research Engineer

Rome (IT)

June 2024 – present

Leonardo

Conducted industrial R&D within Leonardo’s CTO Area / AI Lab across NLP, agentic AI, adversarial machine learning, and edge AI for defence and security applications.

• Led development of agentic AI framework, enabling multi-modal LLM pipelines with RAG, MCP and ReAct.

• Built multi-agent, multi-LLM workflows to optimize prompts and decisions, improving average pass@1 of 8%.

• Deployed edge object-detection models achieving mAP@0.5 = 0.92 on small aerial objects and ~4 FPS on a Variscite i.MX 8M Plus edge device.

• Implemented 10+ adversarial ML attacks, including gradient-free black-box methods, expanding a PoC into a ~€2M project; developed physical patches reducing mAP@0.5 by up to ~70 points on recent YOLO models.

• Building pipelines for knowledge-graph extraction, and RL post-training of small LLMs for multi-turn reasoning.

Artificial Intelligence Engineer

Rome (IT)

Sept 2023 – June 2024

SMART-I

Worked on computer vision and deep learning systems for smart-city applications, focusing on people/vehicle detection and traffic analysis. Deployed solutions on edge devices embedded in smart cameras.

• Trained models for Automatic License Plate Recognition (ALPR), boosted accuracy (mAP) by 46%.

• Optimized C++ camera backend, reducing CPU usage by 30% through asynchronous processing.

• Deployed computer-vision pipelines on edge devices and Google Coral TPUs for real-time inference.

Teaching Assistant

Fisciano (IT)

Mar 2023 – June 2023

University of Salerno

• Assisted OS and CPU Architecture students with MIPS assembly and exercise design.

Research Intern

Enschede (NL)

Sept 2022 – Dec 2022

University of Twente

Joint internship with the University of Salerno, exploring video generation with controllable emotion using GANs.

• Designed a novel GAN loss improving emotion accuracy by 40% while reducing identity drift by ~2%.

• Conducted human evaluation studies with 70% preference over baseline models.

Education

University of SalernoFisciano (IT)

Sept 2020 – Jan 2023

M.Sc. in Computer Engineering

Graduated with 110/110 cum laude (Honors)

University of SalernoFisciano (IT)

Sept 2017 – July 2020

B.Sc. in Computer Engineering

Graduated with 110/110 cum laude (Honors)

Awards

Innovation Awards

Leonardo

Nov 2025

Honorable Mention, “Best Development 2024” for Fproto, a low-code agentic AI framework for rapid pipeline prototyping.

Top Student Honor

University of Salerno

Mar 2023

Recognized among the top students of the previous 5 years in the DIEM department.

Skills

Programming: Python, C++, C, Java, SQL

GenAI: Hugging Face, vLLM, llama.cpp, LangChain, LangGraph, LangFuse, MCP, RAG, ReAct

ML: PyTorch, TensorFlow, OpenCV, NLP, TF Lite, ONNX, Coral TPU, GStreamer

Backend & Systems: FastAPI, Flask, Qdrant, ZeroMQ, Git, Unix/Linux

Languages: English (fluent, C1), Italian (native)