Skip to content
Gaia Cecchi
IT

Research Scholar · University of Siena (DISPOC)

Research that ships.

Researcher in Explainable AI & UX Design. I turn research into working software people can understand, trust, and audit — for cybersecurity, healthcare, and the industrial edge.

Open to roles in applied AI · AI product · UX research

Work ↓gaia170101@gmail.com
  • LLM fine-tuning · prompt engineering · RAG
  • UX research · Design Thinking
  • 4 peer-reviewed publications

Selected work

Research, applied.

Five projects where explainability had to survive contact with production: real partners, real constraints, measured results.

01 · Featured

Cracker Breaker

The AI security product that makes Wazuh SIEM logs readable for everyone: LLMs turn raw alerts into human-readable, actionable intelligence for SMEs.

Cybersecurity · LLM engineering · Explainable AI · Wazuh SIEM

Read the case study →

~90%
false positives filtered out
<1 min
real-time response on critical alerts
3
role-differentiated views

HEART / DiDi

Clinical AI assistant for patient safety, from pre-op hygiene protocols to post-discharge monitoring: RAG over clinical records, privacy-first OCR at the edge, and an interface designed to stimulate — not replace — clinical judgment.

Healthcare AI · RAG · Privacy by design · Human-in-the-loop

+40%

comprehension of out-of-range values

Read the case study →

A2I — Algorithmic Auditability Index

Paper in submission

Regulations demand auditable AI; A2I measures it. A six-dimension index scoring industrial AI pipelines on traceability — validated across predictive maintenance, quality inspection, energy balancing, and supply chain.

AI governance · Auditability · EU AI Act · Applied research

6

auditability dimensions measured

Read the case study →

SAAM — Seco AI Apps Market

Edge AI research with SECO: benchmarking sub-8B LLMs on industrial chipsets, RAG over technical manuals during machine faults, and voice interfaces — all running on-device, where the data stays.

Edge AI · On-device LLMs · Benchmarking · Industrial IoT

<8B

parameters, running fully on-device

Read the case study →

About

UX researcher by training, LLM engineer by practice.

I work at the intersection of two disciplines that rarely share a desk: user experience research and applied large-language-model engineering. My degrees — both summa cum laude, in Communication Sciences and in Experience Design — trained me to start from people; my research position at the University of Siena trained me to end with working systems.

As a Research Scholar at the Department of Social, Political and Cognitive Sciences (DISPOC), I build and evaluate LLM-based systems with industrial partners: translating raw SIEM alerts into decisions non-specialists can act on, benchmarking sub-8B models on edge hardware, and designing retrieval pipelines for clinical data. Explainability is the through-line — not as a compliance checkbox, but as the property that makes AI usable by the people it is meant to serve.

I’m looking for a role where that combination matters: applied AI, AI product, or UX research teams building systems that real people — analysts, clinicians, technicians — have to trust with real decisions.

Gaia Cecchi
Languages
Italian — Native
English — C1/C2 (CEFR)
Siena, Italy
Research Scholar · University of Siena (DISPOC)

Experience & capabilities

Where I’ve worked, what I work with.

Experience

  1. April 2025 – present · Siena, Italy

    Research Scholar

    University of Siena — Department of Social, Political and Cognitive Sciences (DISPOC)

    Applied AI research and development with industrial partners, focused on making LLM-based systems explainable, private, and deployable.

    • Cracker Breaker — AI module of a next-generation security product: LLM integration with the Wazuh SIEM to translate raw alerts into actionable intelligence for non-specialists.
    • SAAM — edge AI with SECO: benchmarking sub-8B LLMs on industrial chipsets and prototyping on-device assistants for privacy-sensitive diagnostics.
    • HEART — clinical AI: RAG over patient records, privacy-first OCR, and human-in-the-loop interface design for a surgical clinic.
    • A2I — co-designed a quantitative index of AI auditability, validated on four industrial case studies (in submission to Computers in Industry).

  2. October 2022 – December 2024 · Siena, Italy

    Social Media Manager

    Taste of Italy Srl

    Ran multi-platform content strategy alongside my studies: trend and pattern analysis, bespoke content aligned to audience profiles, competitor analysis, and continuous iteration on user feedback.

Education

  1. 2022 – 2024 · 110/110 summa cum laude

    MSc, Communication Strategies and Techniques — Technologies and Methods for Experience Design

    University of Siena

    UX and user-centered design with Design Thinking methodology, digital prototyping, Python for data science, and qualitative/quantitative research methods. Thesis: designing an innovation path for Siena’s candidacy as a UNESCO City of Music.

  2. 2019 – 2022 · 110/110 summa cum laude

    BSc, Communication Sciences

    University of Siena

    Communication theory, data collection and analysis, and persuasive writing. Thesis: developing the getting-started mindset for Siena’s first energy community through Design Thinking.

Capabilities

LLM engineering

  • Fine-tuning
  • Prompt engineering (chain-of-thought)
  • RAG pipelines
  • Model benchmarking & evaluation
  • Sub-8B / on-device models
  • Ollama

Machine learning & data

  • Python
  • PyTorch
  • TensorFlow
  • Scikit-learn
  • Pandas
  • NumPy
  • Statistical evaluation

Security & edge

  • Wazuh SIEM/XDR
  • Google Cloud
  • Edge Impulse
  • Edge deployment (Snapdragon X, Meteor Lake, Hailo)
  • NIS2 / EU AI Act awareness

UX & design

  • Design Thinking
  • UX research methods
  • Figma
  • Balsamiq
  • Prototyping

Tools & workflow

  • Git / GitHub
  • Docker
  • FastAPI
  • n8n
  • SQLite

Certifications

Publications

Peer-reviewed, not just prototyped.

The methods behind my projects have survived review: three journal articles (MDPI) and a book chapter.

  • 2026First author

    Connecting the Dots: A Systematic Literature Review of Explainable AI, Cybersecurity, Human-Centered Design and Edge Computing

    Cecchi, G.; Benelli, F.; Caronna, M.; Palma, G.; Rizzo, A. · Journal of Cybersecurity and Privacy (MDPI), 6(3), 91 · Peer-reviewed journal article

    DOI: 10.3390/jcp6030091 (opens in a new tab)

  • 2025

    Leveraging Large Language Models for Scalable and Explainable Cybersecurity Log Analysis

    Palma, G.; Cecchi, G.; Caronna, M.; Rizzo, A. · Journal of Cybersecurity and Privacy (MDPI), 5(3), 55 · Peer-reviewed journal article

    ▸ Local LLMs (qwen2.5:7b, gemma3:4b, llama3.2:3b) reach F1 0.928 on vulnerability detection vs XGBoost 0.555 and LightGBM 0.432

    DOI: 10.3390/jcp5030055 (opens in a new tab)

  • 2025

    Large Language Models for Predictive Maintenance in the Leather Tanning Industry: Multimodal Anomaly Detection in Compressors

    Palma, G.; Cecchi, G.; Rizzo, A. · Electronics (MDPI), 14(10), 2061 · Peer-reviewed journal article

    DOI: 10.3390/electronics14102061 (opens in a new tab)

  • 2024

    Comunicare col pensiero: le Brain-Computer Interfaces

    [Communicating with thought: Brain-Computer Interfaces]

    Cecchi, G.; Masi, L. · Tecnorama Vol. 2: Etica. Tecnologie. Sostenibilità (O. Parlangeli, Ed.), C&P Adver Effigi · Book chapter

Full record on Google Scholar (opens in a new tab) · ORCID (opens in a new tab)

Contact

Let’s talk

I’m open to opportunities in applied AI, AI product, and UX research. The fastest way to reach me is email; I’m also active on LinkedIn.

Email

gaia170101@gmail.com

Elsewhere