root@hub:~/cv$
cat Ludovico_De_Salvo.md
[cd ..]

LUDOVICO DE SALVO

AI Engineer & Management Engineer

SYS.LOC: Rome, ItalySYS.MAIL: ludovicodesalvo@proton.meSYS.NET: linkedin.com/in/ludovicodesalvoSYS.REPO: github.com/LudovicoDeSalvo

## ABOUT_ME

AI & Robotics Master’s student with a solid background in Management Engineering. This dual perspective defines my approach: building systems that work not only mathematically, but that solve real-world problems efficiently. My goal is to bridge the gap between advanced Deep Learning research and practical Business Optimization.

## TOP_SKILLS

Machine LearningDeep LearningReinforcement LearningComputer VisionPyTorchNLP

## PROJECTS

> Bio-Adaptive Music Engine (BAME): Closed-Loop RL System

  • Developed a Deep Reinforcement Learning framework designed to regulate human physiological states (arousal/stress) through music.
  • Treated human physiology as a control problem, engineering a proprietary 'World Model' to simulate biological reactions.
  • Trained an agent to optimize physiological outcomes without real-time human risk.

> End-to-End License Plate Recognition Pipeline

  • Engineered a modular pipeline for real-time license plate detection and recognition on the CCPD2019 dataset.
  • Orchestrated a full end-to-end workflow: Data ingestion -> YOLOv Detection -> Cropping/Alignment -> Recognition -> Inference.
  • Implemented and compared a baseline CNN+LSTM+CTC against a Transformer-based PDLPR (Parallel Decoder) model.

> OCR Post-Correction with Fine-Tuned LLMs

  • Developed a portable pipeline to correct noisy OCR text outputs using Parameter-Efficient Fine-Tuning (PEFT) on LLMs.
  • Implemented LoRA and 8-bit quantization to fine-tune Llama and Minerva models on consumer hardware.
  • Designed an 'LLM-as-a-Judge' automated evaluation system using the Gemini API, validated against human scoring via correlation analysis.

> ReMERT: Enhanced Deep Q-Learning via Experience Replay

  • Built a PyTorch implementation of Deep Q-Networks (DQN) enhanced with custom ReMERT (Replay Memory with End-Related Transitions).
  • Replaced uniform experience replay with a prioritized sampling strategy based on inverse distance to terminal states.
  • Significantly improved sample efficiency and convergence speed in the CartPole-v1 environment.
>./VIEW_ALL_PROJECTS.sh

## EDUCATION

Master Student in Artificial Intelligence & Robotics

[ 2024 - Present ]
Sapienza University, Rome, Italy

English Master Degree. Relevant already completed courseworks: Machine Learning, Deep Learning, Computer Vision, Reinforcement Learning, Neural Networks, Multilingual Natural Language Processing (MNLP).

Bachelor Degree in Management Engineering (L-8)

[ 2017 - 2024 ]
Università della Calabria, Italy

## EXPERIENCE

Waiter

[ 2019 - 2020 ]
“Ristorantino”, Trebisacce, Italy

Developed strong time management skills and ability to work under pressure in a fast-paced environment.

## DETAILED_SKILLS

> Concepts

Artificial IntelligenceDeep LearningReinforcement LearningComputer VisionNatural Language Processing (NLP)Neural NetworksLarge Language Models (LLM)Retrieval-Augmented Generation (RAG)Convolutional Neural Networks (CNN)Long Short-Term Memory (LSTM)Transformer ModelsLLM fine-tuning

> Tools & Systems

LinuxDockerGitCLIJupyterRobot Operating System (ROS)KatharaMicrosoft Office (Word, Excel, PowerPoint)Android (ADB, Custom ROMs, Fastboot)PC Building & Hardware Assembly

> Languages

PythonJavaC++TypeScriptJavaScriptMATLABHTMLCSSPlanning Domain Definition Language (PDDL)

> Frameworks & Libs

PyTorchNumPyMatplotlibOpenCVYOLOFAISSMERTOpenAI GymGoogle Gemini

## LANGUAGES

Italian: Native (C2 level)
English: Full Professional proficiency (C1 level with IELTS certification)
// EOF: Ludovico_De_Salvo.md
Ludovico De Salvo

Contact

ludovicodesalvo.dev/cv
Rome, Italy
ludovicodesalvo@proton.me
linkedin.com/in/ludovicodesalvo
github.com/LudovicoDeSalvo

Top Skills

Machine LearningDeep LearningReinforcement LearningComputer VisionPyTorchNLP

Technical Skills

Concepts

AI, Deep Learning, Reinforcement Learning, Computer Vision, NLP, Large Language Models (LLM), Retrieval-Augmented Generation (RAG), LLM Fine-tuning

Languages

Python, Java, C++, TypeScript, JavaScript, MATLAB, HTML/CSS, PDDL

Frameworks & Libs

PyTorch, NumPy, Matplotlib, OpenCV, YOLO, FAISS, MERT, OpenAI Gym, Google Gemini

Tools & Systems

Linux, Docker, Git, CLI, Jupyter, Robot Operating System (ROS), Kathara, Android ADB/Custom ROMs, PC Hardware Assembly

Languages

Italian: Native (C2)
English: Full Professional (C1, IELTS Certified)

Ludovico De Salvo

AI Engineer & Management Engineer

ludovicodesalvo.dev/cv

Profile Summary

AI & Robotics Master’s student with a solid background in Management Engineering. This dual perspective defines my approach: building systems that work not only mathematically, but solve real-world problems efficiently. My goal is to bridge the gap between advanced Deep Learning research and practical Business Optimization.

Technical Engineering Experience

Bio-Adaptive Music Engine (BAME): Closed-Loop RL System

Developed a Deep Reinforcement Learning framework regulating human physiological states (arousal/stress) through music. Treated human physiology as a control problem, engineering a proprietary "World Model" to simulate biological reactions. Trained an agent to optimize physiological outcomes without real-time human risk.

End-to-End License Plate Recognition Pipeline

Engineered a modular pipeline for real-time license plate detection and recognition on CCPD2019. Orchestrated full workflow: Data Ingestion → YOLOv Detection → Cropping/Alignment → Recognition → Inference. Implemented and compared baseline CNN+LSTM+CTC against a Transformer-based Parallel Decoder (PDLPR) model.

OCR Post-Correction with Fine-Tuned LLMs

Developed a portable pipeline to correct noisy OCR text outputs using PEFT on LLMs. Implemented LoRA and 8-bit quantization to fine-tune Llama and Minerva models on consumer hardware. Designed an "LLM-as-a-Judge" automated evaluation system using the Gemini API, validated against human scoring.

ReMERT: Enhanced DQN via Prioritized Experience Replay

Built a PyTorch implementation of Deep Q-Networks (DQN) enhanced with custom ReMERT (Replay Memory with End-Related Transitions), replacing uniform experience replay with prioritized sampling based on inverse distance to terminal states. Improved sample efficiency in CartPole-v1.

Education

M.Sc. in Artificial Intelligence & Robotics

2024 - Present
Sapienza University, Rome, Italy

English Master Degree. Completed coursework: Machine Learning, Deep Learning, Computer Vision, Reinforcement Learning, Neural Networks, Multilingual NLP.

B.Sc. in Management Engineering

2017 - 2024
Università della Calabria, Rende, Italy

Additional Experience

Waiter

2019 - 2020
“Ristorantino”, Trebisacce, Italy

Developed strong time management skills and ability to work under pressure in a fast-paced environment.