Sign in

Federica Gazzelloni

@fgazzelloni.bsky.social
1.7K followers 131 following 220 posts

Synthetic Data & AI Training | AI Audit, Ethics & Explainability | Human–AI Systems Founder @ DataMetricus | IHME GBD | Author 📘 amzn.to/4swdSDB 🤩👇🏻 Let's get connected! 👇🏻🤩

PostsRepliesMedia
Federica Gazzelloni @fgazzelloni.bsky.social · 16h
Why a high benchmark score doesn't mean what you think: 1️⃣ Scores are relationships, not innate traits. 2️⃣ Goodhart’s Law is real—tests reshape what they measure. 3️⃣ Performance ≠ Learning or generalizability. Continue Reading: federicagazzelloni.substack.com/p/notes-on-d...
federicagazzelloni.substack.com
Notes on Data and Learning — N.30
When Performance Is Not Learning
010
Federica Gazzelloni @fgazzelloni.bsky.social · 28/09/2026
🎥 The recording of our @RLadiesRome workshop, “Introduction to Machine Learning in Epidemiology with R,” is now available! Watch: www.youtube.com/watch?v=v8bx... Code: github.com/fgazzelloni/... #RStats #MachineLearning #Epidemiology
github.com
GitHub - fgazzelloni/Intro-to-ml-in-epidemiology-with-r: Reproducible R-Ladies Rome workshop on machine learning in epidemiology with mlr3 and West Nile virus surveillance data.
Reproducible R-Ladies Rome workshop on machine learning in epidemiology with mlr3 and West Nile virus surveillance data. - fgazzelloni/Intro-to-ml-in-epidemiology-with-r
060
Federica Gazzelloni @fgazzelloni.bsky.social · 18/09/2026
New issue: Notes on Data and Learning — N.29 When a model shapes a decision, the outcome becomes part of the next dataset. The analytical workflow is no longer a straight line, but a feedback loop. Read on Substack: federicagazzelloni.substack.com/p/notes-on-d... #DataScience #AI #ActuarialScience
federicagazzelloni.substack.com
Notes on Data and Learning — N.29
When the Model Changes the Data
000
Federica Gazzelloni @fgazzelloni.bsky.social · 13/09/2026
New issue out! 📊 Notes on Data and Learning — N.28: Before the Pipeline, There Is a Question AI can retrieve data, write code, transform datasets, build models, and explain results. Read more federicagazzelloni.substack.com #DataScience #AI #DataAnalytics #RStats
federicagazzelloni.substack.com
Federica’s Substack | Substack
I share curated links to books, software-related resources, and tools that I personally use or consider useful for data analysis, statistical work, programming, and technical productivity. These…
030
Federica Gazzelloni @fgazzelloni.bsky.social · 09/09/2026
📢 New R-Ladies Rome workshop! Introduction to Machine Learning in Epidemiology with R RSVP: luma.com/xmchetzk A practical session. No previous ML experience required. Everyone is welcome! 💜 #RLadies #RLadiesRome #RStats #MachineLearning #Epidemiology
041
Federica Gazzelloni @fgazzelloni.bsky.social · 02/09/2026
From reproducibility to auditability in AI-augmented analytics When AI enters an analytical workflow, preserving code is no longer enough. We also need to reconstruct context. Continue reading: federicagazzelloni.substack.com/p/notes-on-d... #DataScience #AI #ReproducibleResearch #AIGovernance
federicagazzelloni.substack.com
Notes on Data and Learning — N.27
When Reproducibility Becomes Governance
120
Federica Gazzelloni @fgazzelloni.bsky.social · 21/08/2026
The latest issue of Notes on Data and Learning is out: When Data Becomes Conversational. Read the full issue: federicagazzelloni.substack.com/p/notes-on-d... #DataScience #AI #ReproducibleResearch
federicagazzelloni.substack.com
Notes on Data and Learning - N.26
When Data Becomes Conversational
000
Federica Gazzelloni @fgazzelloni.bsky.social · 05/08/2026
📢 Next up at R-Ladies Rome! Typst for Efficient Typesetting with Derek Sollberger (@Princeton) Discover how Typst + Quarto, publication-quality PDFs—without the complexity of LaTeX. 📅 Registration: luma.com/k9bw4g7e #RLadiesRome #RStats #Typst #Quarto #Reproducible
luma.com
Typst for Efficient Typesetting · Zoom · Luma
Creating professional-looking PDFs for articles, reports, lecture notes, and books has traditionally meant learning LaTeX. Today, Typst offers a modern…
140
Federica Gazzelloni @fgazzelloni.bsky.social · 07/07/2026
📚 New issue out! Notes on Data and Learning — N.25 This week I reflect on something I hadn’t fully appreciated before: the real cost of modern technical work is often not multitasking, but rebuilding context. 👇 Read the full issue: federicagazzelloni.substack.com
federicagazzelloni.substack.com
Federica’s Substack | Substack
I share curated links to books, software-related resources, and tools that I personally use or consider useful for data analysis, statistical work, programming, and technical productivity. These…
010
Federica Gazzelloni @fgazzelloni.bsky.social · 26/06/2026
📚 New issue — Notes on Data and Learning N.24 Learning R isn't just memorizing functions. With AI automating code, the crucial skill is understanding workflows — how data, code, docs, and decisions connect. Read: federicagazzelloni.substack.com #RStats #DataScience #AI
040
Federica Gazzelloni @fgazzelloni.bsky.social · 17/06/2026
New issue out ✍️ Notes on Data and Learning — N.23 When Priorities Choose Themselves This week was supposed to be about organization. Continue Reading: federicagazzelloni.substack.com
federicagazzelloni.substack.com
Federica’s Substack | Substack
I share curated links to books, software-related resources, and tools that I personally use or consider useful for data analysis, statistical work, programming, and technical productivity. These…
000
Federica Gazzelloni @fgazzelloni.bsky.social · 10/06/2026
📢 Upcoming R-Ladies Rome Workshop From Data Analysis to Publication: Reproducible Research with R and Quarto 🎙️ With Betsabé Cohen & Jesica Formoso (R-Ladies+ Buenos Aires) 📅 Registration: luma.com/j2qxv1pr #RLadiesRome #RStats #QuartoPub #ReproducibleResearch
luma.com
From data analysis to publication: Reproducible research with R and Quarto · Zoom · Luma
Research is most valuable when others can understand it, reproduce it, and build upon it. Yet many research workflows still rely on disconnected files, manual…
010
Federica Gazzelloni @fgazzelloni.bsky.social · 07/06/2026
In this week’s issue, I reflect on memory, judgment, programming education,. Read the full issue : federicagazzelloni.substack.com/p/notes-on-d... #RStats #DataScience #ArtificialIntelligence #Programming #Education #Teaching #Learning #ComputationalThinking #Statistics #NotesOnDataAndLearning #AI
federicagazzelloni.substack.com
Notes on Data and Learning — N.22
Learning R in the Age of AI
061
Federica Gazzelloni @fgazzelloni.bsky.social · 25/05/2026
New issue published. Notes on Data and Learning — N.20 Learning Structure: Notes, Graphs, and the Coordination of Uncertainty This week I wrote about Obsidian and Ebola. How much structure is enough to preserve reasoning without creating noise? Read More: federicagazzelloni.substack.com
federicagazzelloni.substack.com
Federica’s Substack | Substack
I share curated links to books, software-related resources, and tools that I personally use or consider useful for data analysis, statistical work, programming, and technical productivity. These…
010
Federica Gazzelloni @fgazzelloni.bsky.social · 17/05/2026
New issue of Notes on Data and Learning published: N.19 — The Geometry of Coordination This week’s reflection started from an unexpected place: the strange way ideas continue moving after a project is technically “finished.” Continue Reading federicagazzelloni.substack.com
federicagazzelloni.substack.com
Federica’s Substack | Substack
I share curated links to books, software-related resources, and tools that I personally use or consider useful for data analysis, statistical work, programming, and technical productivity. These…
020
Federica Gazzelloni @fgazzelloni.bsky.social · 13/05/2026
🦠 Questa Domenica terrò un incontro online in italiano su Hantavirus, epidemiologia e diffusione delle infezioni negli ambienti chiusi e connessi. 🎁 Giveaway di 1–2 copie del mio libro. 🔗 luma.com/rhx85w27 📅 Domenica 🕡 17:00 💻 Online #Hantavirus #SalutePubblica
luma.com
Malattie Infettive nel Mondo che Cambia - Hantavirus · Zoom · Luma
🦠 Malattie Infettive nel Mondo che Cambia Incontro online con Federica Gazzelloni Negli ultimi giorni il cluster di Hantavirus collegato alla nave MV Hondius…
000
Federica Gazzelloni @fgazzelloni.bsky.social · 10/05/2026
New R-Hack published: Your Script Is Not a Workflow The problem often appears only later, when the session is restarted and the analysis depends on objects, steps, or assumptions that were never made explicit. 👉 Read the full issue here: romerusersgroup.github.io/content/rhacks
romerusersgroup.github.io
R-Hacks – Rome R-Users Group
Each post focuses on a small, practical task you can apply immediately—whether you’re cleaning data, visualising results, automating a workflow, or debugging a stubborn function. Think of it as a…
121
Federica Gazzelloni @fgazzelloni.bsky.social · 10/05/2026
New issue published: Notes on Data and Learning — N.18 From Workflows to Coordination: The Power of Direct Access AI responds to structure. Humans respond to meaning, emotion, timing, and willingness. Read More: federicagazzelloni.substack.com #AI #MCP #LLM #AgenticAI #HumanAI
federicagazzelloni.substack.com
Federica’s Substack | Substack
I share curated links to books, software-related resources, and tools that I personally use or consider useful for data analysis, statistical work, programming, and technical productivity. These…
150
Federica Gazzelloni @fgazzelloni.bsky.social · 06/05/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 05/05/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
120
Federica Gazzelloni @fgazzelloni.bsky.social · 04/05/2026
New R-Hack published: Stop rewriting the same code A lot of R workflows don’t become difficult because they are complex. AI makes it easy to generate code. It also makes it easy to repeat the same logic over and over. 👉 romerusersgroup.github.io #rstats
romerusersgroup.github.io
Rome R-Users Group
Alert! Free tutorials available!
120
Federica Gazzelloni @fgazzelloni.bsky.social · 04/05/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 03/05/2026
Issue N.17 — Learning Control Getting a system to work is not the hard part. Keeping it under control as conditions change is. AI follows instructions. Humans don’t always. → federicagazzelloni.substack.com
federicagazzelloni.substack.com
Federica’s Substack | Substack
I share curated links to books, software-related resources, and tools that I personally use or consider useful for data analysis, statistical work, programming, and technical productivity. These…
110
Federica Gazzelloni @fgazzelloni.bsky.social · 03/05/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 02/05/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
110
Federica Gazzelloni @fgazzelloni.bsky.social · 01/05/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 30/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 29/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 28/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
010
Federica Gazzelloni @fgazzelloni.bsky.social · 27/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 26/04/2026
New R-Hack: Start learning R faster Most people try to learn everything at once. That’s where it slows down. Start with one workflow. Repeat it. Build from there. 👉 romerusersgroup.github.io/content/rhac... #rstats #DataScience
romerusersgroup.github.io
R-Hacks – Rome R-Users Group
Each post focuses on a small, practical task you can apply immediately—whether you’re cleaning data, visualising results, automating a workflow, or debugging a stubborn function. Think of it as a…
0132
Federica Gazzelloni @fgazzelloni.bsky.social · 26/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
020
Federica Gazzelloni @fgazzelloni.bsky.social · 25/04/2026
Issue N.16 — From Prompts to Context Same prompt. Different outputs. The difference was context. AI is shifting from prompt design → context design. Read → federicagazzelloni.substack.com
federicagazzelloni.substack.com
Federica’s Substack | Substack
I share curated links to books, software-related resources, and tools that I personally use or consider useful for data analysis, statistical work, programming, and technical productivity. These…
000
Federica Gazzelloni @fgazzelloni.bsky.social · 25/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 24/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
010
Federica Gazzelloni @fgazzelloni.bsky.social · 23/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 22/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 22/04/2026
🔥 Reminder: our next @RLadiesRome session is coming up “The Roots of AI: NLP and N-gram Models in R” We’ll build a next-word prediction model from scratch Data → counts → probabilities → prediction See you there 👇 www.meetup.com/rladies-rome...
meetup.com
The roots of AI: Natural Language Processing & N-gram models, Thu, Apr 23, 2026, 7:00 PM | Meetup
Artificial Intelligence often feels like a black box, but many of its core ideas come from simple and elegant statistical concepts. In this session, we will step back from
001
Federica Gazzelloni @fgazzelloni.bsky.social · 21/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Reposted by Federica Gazzelloni
n8n @n8n.io · 08/04/2026
Get tips, practical patterns, & downloadable templates showing how to mix deterministic steps and AI steps for workflows that are faster, less expensive, and more reliable. 📚 bit.ly/41LeNDY
031
Federica Gazzelloni @fgazzelloni.bsky.social · 20/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 20/04/2026
Issue N.15 — Learning to Prioritise Working more ≠ progressing more. The real constraint is attention, not time. Starting is easy. Staying with one task long enough to make it coherent is not. Read → federicagazzelloni.substack.com
federicagazzelloni.substack.com
Federica’s Substack | Substack
I share curated links to books, software-related resources, and tools that I personally use or consider useful for data analysis, statistical work, programming, and technical productivity. These…
010
Federica Gazzelloni @fgazzelloni.bsky.social · 20/04/2026
New R-Hack: Stop guessing why your code breaks Most people don’t debug. They regenerate. Small shift: break the pipeline → find the failure Read: 👉 romerusersgroup.github.io/content/rhac... #rstats #datascience #AI
romerusersgroup.github.io
R-Hacks – Rome R-Users Group
Each post focuses on a small, practical task you can apply immediately—whether you’re cleaning data, visualising results, automating a workflow, or debugging a stubborn function. Think of it as a…
130
Federica Gazzelloni @fgazzelloni.bsky.social · 19/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 18/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 17/04/2026
Quickly create reports & dashboards with Quarto & R Markdown. Datametricus offers powerful tools for data exploration and storytelling. #Quarto #RMarkdown #DataAnalysis #DataVisualization
datametricus.com
Quarto and R Markdown integration for data analysis & visualization.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 17/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 16/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
110
Federica Gazzelloni @fgazzelloni.bsky.social · 15/04/2026
Dashboards built before models are hypotheses dressed as findings. #HealthData #Epidemiology #DataScience #ReproducibleResearch #PublicHealth
Dashboards built before models are hypotheses dressed as findings.
000
Federica Gazzelloni @fgazzelloni.bsky.social · 15/04/2026
Datametricus provides data visualization & analysis services. They help businesses understand data effectively with reports and dashboards. #DataVisualization #Analytics #Datametricus
datametricus.com
Datametricus offers data visualization and analysis services, helping businesses make data-driven decisions. They provide tools and expertise for creating insightful reports and dashboards.
000