I build the data layer, then put models on top of it
I am Rahul Singh. VizzingData is my independent data and AI practice, based in Hyderabad and working with clients across time zones. I do the work myself — there is no bench, no account manager, and nobody you have not met.
How this practice got here
This site began as a Tableau tutorial blog, and that is not an accident of history — it is the foundation the rest sits on. Years of building dashboards and reporting for other people teaches you something that no amount of model expertise substitutes for: where an organisation’s real data lives, why nobody can join it, and which numbers people actually make decisions on.
That turns out to be exactly the problem large language models run into. A model is only as useful as the data you can put in front of it, and most organisations cannot put their own data in front of anything. So the work moved naturally from reporting into applied AI — same data layer, higher ceiling. I did not leave BI behind to chase AI. The BI is why the AI works.
What I have actually shipped
Running systems, not pilots. Client work is anonymised; the rest is my own.
A production Claude application over operational data
For a teleradiology client: a consolidated data layer of around 424,000 records feeding three automated reporting cadences, an anomaly detector built on robust statistics, and a natural-language interface over the whole thing. Prompt caching and turn limits cut inference cost by roughly 48%.
A fully automated daily content service
Live since June 2026, generating and publishing on a daily and weekly schedule with no manual step. Unglamorous and the best possible proof that something keeps running when nobody is watching it.
An education product with cost-tiered models
Generates teaching worksheets, using a cheaper model for generation and a stronger one for the steps that need judgment. The split cut cost per run by roughly two thirds without a quality drop that mattered.
A multi-agent content pipeline
A chain of specialised agents handing work to each other through a defined sequence. Built mostly to learn where multi-agent designs genuinely beat a single well-prompted call — and where they are just expensive.
A book-lending progressive web app
Real users, real data, in production. Built because consumer software teaches you things enterprise reporting never will — mostly about everything that goes wrong when real people touch your system.
A Tableau tutorial library and public portfolio
Years of tutorials on this site covering Tableau concepts, charts, calculations and advanced techniques, plus published vizzes and data journalism. All of it still free, and still the clearest demonstration of the BI foundation.
My résumé, built in Tableau
An interactive version of my background, published to Tableau Public — because a résumé built in the tool is a better demonstration than a line on one.
What I work with
- Python
- SQL
- Power BI
- Tableau
- BigQuery
- Claude API
- Prompt caching & cost control
- Retrieval over client data
- Agent and tool-use design
- Evaluation
- Azure (App Service, Foundry)
- AWS Bedrock
- Google Cloud
- Firebase
- Robust anomaly detection
- Forecasting
- Backtesting & tuning
- SLA and breach analysis
How I think about this work
Most AI projects do not fail on the model. They fail because the data was never accessible, because nobody costed the inference, or because the output could not be trusted enough for anyone to act on it. Those are engineering problems with known answers, and they are the three I spend my time on.
So I will tell you when something is not worth building. A smaller system that people actually use beats an ambitious one that quietly gets switched off, and I would rather lose the larger piece of work than deliver the second thing.
Certifications
Twenty Tableau credentials, including the full set of role-based certifications. Verifiable on Credly.
Elsewhere
Looking for help with something?
The services page covers what I do and what it costs to run. If you already know what you are trying to see, write to me directly and I will tell you honestly whether it is worth building.
rahul@vizzingdata.com