A free, hands-on data-literacy lab for people who decide with data but don't code. Touch real data, watch a wrong conclusion form, and learn the judgment behind every chart, model, and AI answer.
Data-Lab is a free, open-source data-literacy platform containing 68 hands-on interactive demos that teach the complete data lifecycle — collecting, cleaning, filtering, analysing, engineering, modelling, and governing data — to people who will never write code. It runs in a web browser, requires no account, no credit card and no programming, and is fully bilingual in English and Arabic.
Every demo is operated with sliders, dropdowns and buttons rather than a code editor, and each one ends with a plain-language “what this means for a leader” takeaway. The demos follow a single running story — a small fictional shop called Nour Store — from raw, messy order data all the way to a business decision, so the material accumulates into one narrative instead of scattered exercises.
Data-Lab was built by Eng. Mulham Fetna as the interactive backbone of the From Data to Decisions workshop by Neurobotics Academy. It is released under the AGPL-3.0 licence and archived on Zenodo with a permanent DOI, so it can be cited in academic work, forked by other educators, and redistributed — provided derivatives stay open and keep attribution.
The distinguishing idea is judgment over construction. Most data education teaches you to build data products. Data-Lab teaches you to judge them: to recognise when a chart is lying, when a model's accuracy is an artefact, when a sample is biased, and when an AI answer is invented. Those are the skills a person who funds, approves, buys and governs data work actually needs.
We surveyed around 35 platforms — paid coding schools, free hands-on sites, and executive programmes. They split into three camps, and each one structurally shuts out the leader who won't write code.
DataCamp, Kaggle Learn, Dataquest, Google Data Analytics. Genuinely interactive — but you must write code. A wall for a non-programmer.
Tableau, Qlik, “AI for Everyone”, MIT Sloan, Wharton. Right audience — but you watch; you never touch a dataset and see it lie.
Free tiers stop at chapter one; audit windows shrink to one or two weeks; the interesting parts want an account or a card.
No one lets a non-coder manipulate real data and models — for free, with no login, in Arabic. That is exactly what Data-Lab does.
Scored against the nearest alternative in each camp. The sharpest single edge is Arabic — the entire hands-on tier of data education is English-only.
| Brilliant | Tableau / Qlik | Kaggle / DataCamp | Elements of AI | Data-Lab | |
|---|---|---|---|---|---|
| Free, with real depth | ✕ | ✓ | ✕ | ✓ | ✓ |
| No signup / no card | ✕ | ✕ | ✕ | ~ | ✓ |
| Hands-on & interactive | ✓ | ✕ | ✓ | ~ | ✓ |
| No coding required | ✓ | ✓ | ✕ | ✓ | ✓ |
| Judgment-first, leader-framed | ✕ | ~ | ✕ | ~ | ✓ |
| Bilingual EN + Arabic | ✕ | ✕ | ✕ | ✓ | ✓ |
| Open-source / forkable | ✕ | ✕ | ✕ | ✕ | ✓ |
| Citable (DOI) | ✕ | ✕ | ✕ | ✕ | ✓ |
Everything runs offline, nothing asks you to write code, and each demo ends with a leader takeaway.
Pick your role and you get a short, curated playlist inside the app — you don't need all 68 demos, you need the six that match your decisions.
Work through Predict and Feature Attribution to see how an impressive accuracy number can be an artefact of how the target was defined — then ask the vendor for the attribution chart.
Generate synthetic records that preserve statistical structure, mask personal identifiers, and check for sampling bias before a result becomes a claim.
Run an A/B test demo, see how confidence intervals shrink with sample size, and learn why testing twenty things and reporting the winner is p-hacking.
Toggle a truncated axis and watch a 4% rise look like a rocket; split the data and watch the trend reverse. Then rebuild the honest version.
Use presenter mode with timers and talk-tracks, project the QR page so the room follows on phones, and finish with the scored capstone. Fork it and adapt it for your own curriculum.
Walk the full lifecycle in an afternoon, see what data work actually involves day to day, and use the learning roadmap to decide what to study next.
Open the hosted app and start clicking. No account, no download. This is the fastest path and the one to project in a workshop.
https://mulhamfetna-datalab.streamlit.app/
Requires Python 3.11 or newer. Everything works offline once installed — useful for classrooms with unreliable internet, and required for the region-restricted scraping demos.
git clone https://github.com/mulhamfetna/Data-Lab.git cd Data-Lab pip install -r requirements.txt ./run.sh # or: streamlit run Home.py
The lab opens at http://localhost:8501. Streamlit also prints a
Network URL — anyone on the same Wi-Fi can open that on their phone, which is the simplest way to
run a room of 30 people with no internet at all.
For self-hosting behind your own domain, or for a reproducible classroom image.
docker compose up -d # → http://localhost:8501
The generative-AI demos work without any key, using a clearly labelled offline simulation. To have them
call a real model, set one free provider key as an environment variable or in
.streamlit/secrets.toml:
GROQ_API_KEY = "gsk_..." # Groq — generous free tier OPENROUTER_API_KEY = "sk-or-..." # OpenRouter — free models HF_TOKEN = "hf_..." # Hugging Face Inference Providers OLLAMA_HOST = "http://localhost:11434" # fully local, fully free
The app auto-detects whichever is present and always displays which provider produced an answer, so a simulation is never mistaken for a real model.
Turn on presenter mode for per-segment timers and talk-track notes, hand out the QR page so participants open the lab on their own phones, pick a role track matching your audience, and close with the scored capstone. Quizzes after each topic make the lessons stick.
Data-Lab is actively developed and will keep growing into a complete, free data-education system. These are commitments, not hopes: the licence guarantees each stage stays open, and every release is archived with its own DOI.
68 demos, bilingual, tested, archived with a DOI.
Written explanations wrapped around every demo.
Short walkthroughs embedded beside each lesson.
Accounts, saved progress, downloadable certificates.
Courses and cohorts on open standards (LTI · xAPI).
Planned additions include deeper written lessons for every module, recorded video walkthroughs, learner accounts with saved progress and certificates, more real-world case studies, expanded Arabic content, and eventually integration with an open learning-management system so schools and NGOs can run structured cohorts. Requests and contributions are welcome on GitHub.
pip install -r requirements.txt, then ./run.sh (or streamlit run Home.py). The lab opens at localhost:8501 and works fully offline. A Dockerfile and docker-compose.yml are included for self-hosting.10.5281/zenodo.21427808, which always resolves to the latest release. The repository contains a CITATION.cff file, and GitHub renders a “Cite this repository” button from it. Each individual release also receives its own version DOI on Zenodo.Built to serve the greater good — free, hands-on data literacy for underserved science and engineering communities, in Arabic and English.
AGPL-3.0 plus a Zenodo DOI means anyone may run, fork and build on Data-Lab — but every derivative must stay open and cite the author. Maximally free and maximally attributed, at once.
@software{fetna_datalab_2026,
author = {Fetna, Mulham},
title = {Data to Decisions: An Interactive
Data-Literacy Workshop Platform},
year = {2026},
publisher = {Zenodo},
doi = {10.5281/zenodo.21427808},
url = {https://datalab.mulhamfetna.com}
}
تعلّم ألّا تنخدع بالبيانات — مجّانًا، بالعربية والإنجليزية.