Skip to main content

Hi, I'm

Andrea Alberti

GenAI Engineer, AI Systems Architect, Multi-Agent Developer

Building intelligent systems with AI

Explore ProjectsGet in touch
  • Machine Learning
  • Multi-Agent Systems
  • LLM Models
  • Deep Learning
  • RAG Systems
Scroll to explore
  • Multi-Agent Systems
  • RAG
  • Vertex AI
  • Gemini
  • Google ADK
  • Dialogflow CX
  • LangGraph
  • Cloud Run
  • BigQuery
  • Agent Engine

01About Me

Who I Am

Graduated with a double degree in Management and Computer Science–Data Science, I have gained substantial experience in multidisciplinary projects. I specialize in applying machine learning, deep learning, and most recently Generative AI techniques to develop innovative and automated solutions.

My academic journey, from Management Engineering (110/110 cum laude) to a Master's in Data Science (110/110 cum laude), has given me a blend of technical expertise and strategic thinking. This allows me to approach complex problems from both a business and a technological perspective.

Professionally, I focus on building intelligent systems—from multi-agent architectures that automate complex business processes to advanced conversational AI—to drive efficiency and reduce operational costs. My goal is to continuously improve myself, using AI to create tangible value.

Education

  1. 2022 — 2024·110/110 L

    MSc. Data Science - Computer Engineering

    University of Pavia

    Download Thesis

  2. 2018 — 2021·110/110 L

    BSc. Management Engineering

    University of Brescia

    Download Thesis

Certifications

Google Cloud Professional ML EngineerGoogle CloudGoogle Cloud Professional Data EngineerGoogle Cloud
Technical Skills48

Core Skills

RAGMulti-Agent SystemsPrompt EngineeringFew-Shot LearningProblem SolvingSystem DesignData AnalysisResearchPublic SpeakingPresentation DesignWorkshop Creation

AI Tools

Claude CodeCodexGitHub CopilotGeminiChatGPTClaudeNotebookLMGemini CLI

Cloud & Infrastructure

Google Cloud PlatformVertex AICloud RunCloud FunctionsBigQueryWebSocketAgent Engine

Programming Languages

PythonSQLHTML/CSS

Frameworks & Libraries

Google ADKDialogflow CXGemini SDKVertex AI SearchTensorFlowPyTorchScikit-learnKerasXGBoostFastAPIFlaskPySparkHadoopMongoDBNetworkXPandasNumPyOpenCVMatplotlib

02Professional Work

Professional Projects

A selection of professional projects where I applied Generative AI and Machine Learning to solve real-world problems.

LiveOct 2024 - Present

AI-Powered Tyre Selection Assistant

Engineered and deployed a complex, production-grade multi-agent conversational system live on the public website of a global tyre industry leader. The architecture guides customers from discovery to purchase via proactive model matching, side-by-side comparisons of proprietary tyre technologies, and direct conversion (dealer locator and online purchase links). Built and migrated to Google Agent Development Kit (ADK) with RAG and external APIs, supported by a resilient, low-latency cross-cloud AWS-GCP integration.Read more
  • Google ADK
  • Multi-Agent
  • RAG
  • +2
CompletedAug 2024 - Oct 2024

Automated Document Validation System

Full-stack implementation of a multi-agent system to automate verification and validation of documents for public funding requests. Developed backend in Python and frontend in HTML, CSS, and JavaScript. Designed a modular and flexible architecture that allows agent behavior adaptation without code modification.Read more
  • Multi-Agent
  • Python
  • Process Automation
  • GCP
  • Full-Stack
CompletedJun 2024 - Sep 2024

Advanced Knowledge Base Chatbot

Implementation of advanced chatbot agent architecture to answer user questions using a knowledge base of web pages and PDF documents. System designed with a main orchestrator agent routing requests to five specialized sub-agents. Developed data management pipeline including PDF parsing, chunking strategy, and security layer for inappropriate questions.Read more
  • Dialogflow CX
  • GCP
  • Multi-Agent
  • RAG
  • Security
  • +1
CompletedJul 2024

Real-Time Multimodal Agent

Development and implementation of real-time multimodal conversational agent based on Google Agent Development Kit (ADK). System capable of processing audio and video inputs simultaneously, sustaining fluid conversations, and autonomously performing browser operations through reasoning and tools. Asynchronous dual-server architecture with WebSocket communication for low-latency bidirectional streaming.Read more
  • Google ADK
  • Gemini Live API
  • Multimodal AI
  • +3
CompletedJul 2024

Luxury Yacht Virtual Assistant

Implementation of virtual assistant with aim of answering user questions based on corporate knowledge base. Configured as RAG (Retrieval-Augmented Generation) application, leveraging Google Cloud ADK and Vertex AI Search to retrieve relevant information from dedicated datastore. Architecture uses Gemini models to process requests and generate accurate, contextualized responses in Italian.Read more
  • RAG
  • Vertex AI Search
  • Google ADK
  • +3
CompletedMay 2024 - Jul 2024

Multi-Agent Ticketing System

Implementation of multi-agent ticketing system to automate user responses. Project developed as Proof of Concept (POC), involving creation of specialized agents, each capable of interacting with external databases via APIs to provide accurate and comprehensive replies. Multi-agent architecture ensures user requests are routed to most competent agent, reducing staff workload and management costs.Read more
  • Multi-Agent
  • Google ADK
  • API Integration
  • +2
CompletedFeb 2024 - Apr 2024

LLM-Based Email Classification Pipeline

Within broader project aimed at automating manual process and reducing operational costs, contribution focused on creating LLM-based pipeline for classification and dispatch of certified emails (PEC). Key activities included prompt engineering and few-shot learning to refine model outputs, along with development of metrics and analytical tools for system performance evaluation.Read more
  • LLM
  • Prompt Engineering
  • Few-Shot Learning
  • Process Automation
  • NLP
  • +1
CompletedInternal Demo

Insurance Liquidation Engine - Demo Project

Developed comprehensive multi-agent collaborative architecture for internal Sharing Days demonstration. System automatically analyzes, processes, and issues liquidation judgments on insurance claims. Architecture combines parallel agents for document analysis with sequential agents for final evaluation, utilizing custom tools, built-in Google Cloud services, and MCP integration.Read more
  • Google ADK
  • Multi-Agent
  • Cloud Run
  • MCP
  • Sequential Agents
  • +1

03Academic Research

Research & Academic Work

8
papers
80+
pages
6
cum laude
  1. Machine Learning

    Heart Disease Detection from Audio Signals

    Advanced Biomedical Machine Learning

    Key Results
    0.82F1-Score
    0.96ROC-AUC
    43.4%TPR @1% FPR
    74.3%TPR @5% FPR
    86.6%TPR @10% FPR
    95.8%TPR @20% FPR
    Jul 202417 pages
    • Scikit-learn
    • Torchaudio
    • Librosa
    • +12
    View DetailsPaperCode
  2. Graph ML

    Disease Prediction with Graph Machine Learning

    Financial Data Science

    Key Results
    Logistic RegressionBest Model
    28%Feature Reduction
    1.5%Accuracy Drop
    9.4%Training Time Reduction
    Dec 202323 pages
    • Python
    • NetworkX
    • Scikit-learn
    • +4
    View DetailsPaperCode
  3. Big Data

    Review Helpfulness Prediction with Big Data

    Data Science & Big Data Analytics

    Key Results
    Random ForestBest Model
    0.026MSE
    0.25R²
    Sep 20236 pages
    • Hadoop HDFS
    • Apache Spark
    • PySpark
    • +7
    View DetailsPaperCode
  4. NLP

    Clickbait Detection in News Headlines

    Machine Learning

    Key Results
    97.12%Accuracy
    0.0%Best FPR
    84.00%FPR Accuracy
    8000 wordsVocabulary
    Feb 20246 pages
    • Python
    • Scikit-learn
    • NumPy
    • +2
    View DetailsPaperCode
  5. Security

    DDoS Attack Detection and Mitigation

    Enterprise Digital Infrastructure

    Key Results
    1.46AF A
    4.14AF MX
    4.46AF NS
    7.30AF ANY
    ≈44 ms @10k ppsMX Mean Query Time
    ≈96 ms @10k ppsNS Mean Query Time
    Jan 202411 pages
    • Python
    • Scapy
    • Wireshark
    • +3
    View DetailsPaperCode
  6. Machine Learning

    Cake Classification Features Analysis

    Machine Learning

    Key Results
    90%Accuracy
    90%Neural Features
    31%Low-Level Features
    80%Transfer Learning
    Feb 20245 pages
    • Python
    • pvml library
    • PVMLNet
    • +3
    View DetailsPaperCode
  7. Computer Vision

    Vanishing Points Detection in Images

    Computer Vision

    Key Results
    500RANSAC Iterations
    5 pixelsThreshold
    10 longest per runHough Lines
    Mar 20247 pages
    • Python 3.9+
    • OpenCV
    • NumPy
    • +3
    View DetailsPaperCode
  8. NLP

    Sentiment Analysis on Social Media

    Machine Learning

    Key Results
    85.4%Accuracy
    82.6%MNB Accuracy
    85.4%LR Accuracy
    1000 wordsBest Vocabulary
    Feb 20245 pages
    • Python
    • Scikit-learn
    • NLTK
    • +2
    View DetailsPaperCode
  • Machine Learning
  • Deep Learning
  • Computer Vision
  • NLP
  • Big Data
  • Graph ML
  • PyTorch
  • TensorFlow
  • PySpark
  • Scikit-learn

04Personal Projects

Personal Projects

Turning ideas into reality through code and creativity.

01 / active

Helmo

An AI financial co-pilot designed to track expenses, simulate purchase scenarios, and plan savings goals. Features an interactive scenario simulator, real-time Safe-to-Spend calculation, and proactive insights via in-app chat and Telegram. Currently in *beta*.

  • Personal Finance
  • AI Assistant
  • Scenario Simulator
  • Safe-to-Spend
  • Telegram
Visit Website
helmofinance.com
Helmo

02 / completed

Liberti Hub

A knowledge hub designed to make Data Science and AI concepts accessible and intuitive. Covers Machine Learning, Deep Learning, Big Data, and Statistical Learning with clear explanations and practical examples.

  • Machine Learning
  • Deep Learning
  • Big Data
  • Education
Visit Website
libertihub.altervista.org
Liberti Hub

03 / active

In Cucina con Chiara

An Italian cooking blog offering a diverse range of recipes. Features include a meal planner, recipe converter, and an AI-powered culinary assistant.

  • Next.js
  • AI Assistant
  • Culinary
  • Web App
  • React
Visit Website
incucinaconchiara.com
In Cucina con Chiara

05My Passions

Beyond Work

Other things I love to do

Doing Sports

Active Body, Active Mind

It is important for me to have moments to take care of my health. I set up a small home gym and I also love playing tennis and football.

Watching Formula 1

The Perfect Mix of Passion, Ability, and Strategy

I love Formula 1 because it represents the pinnacle of racing, combining cutting-edge technology with human skill and strategic thinking.

If you no longer go for a gap which exists, you are no longer a racing driver. — Ayrton Senna

Watching Football

A Long-Lasting Family Tradition

I inherited from my grandfather the passion for football and tennis. I love watching and analyzing matches and I'm a big fan of AC Milan since ever.

06Get in Touch

Let's Connect

Open to opportunities and collaborations. Feel free to reach out!

Email

andalberti99@gmail.com

Available for AI projects · Low latency