Learn Data Science with AI: Full DASHBOARD in 5 MINs with FREE RESOURCES
This tutorial demonstrates how to build a fully functional, interactive data dashboard in 5 minutes using only free resources — Kaggle for datasets and Claude AI for code generation. The presenter walks through selecting a relevant dataset, crafting an effective AI prompt, and generating a deployable HTML dashboard without writing any code manually.
Summary
The video is a step-by-step Hindi-language tutorial on building an interactive data dashboard in under 5 minutes using entirely free tools. The presenter begins by emphasizing the importance of choosing the right dataset from Kaggle (kaggle.com), pointing out that the platform hosts datasets covering customer shopping behavior, YouTube trends, health, finance, and more.
A key emphasis is placed on selecting a dataset that is relevant to the industry you want to apply to. For example, someone targeting e-commerce companies like Amazon or Flipkart should use customer review datasets, while someone targeting finance firms like Zerodha should analyze stock or tariff data. For this tutorial, the presenter picks a child illiteracy dataset, framing it from the perspective of a public policy consultancy firm.
The second major step involves using Claude AI (claude.ai) to generate both the optimal prompt and then the dashboard itself. The presenter teaches a structured prompting approach: first define Claude's role (e.g., 'elite prompt engineer and world-class dashboard designer for public policy and government consulting'), then provide context about who you are and who the audience is (e.g., ministers and development agencies), then state the goal (driving actionable policy decisions), and finally instruct Claude to write an extremely detailed, ready-to-use prompt that will be copy-pasted back to generate the dashboard.
The desired dashboard specifications fed into the prompt include strong interactivity features like slicers, filters, cross-filtering, drill-downs, tooltips, coordinated charts, KPIs, equity analysis sections, strategic recommendations, and a professional government-friendly design with defined color schemes.
The resulting dashboard includes region and income-group dropdowns, a time period selector, country search functionality, a reset button, KPI highlights, hover-based country detail tooltips, trend analysis charts, country-level deep dives, and a strategic recommendations section. The presenter demonstrates filtering by Brazil to show country-specific data isolation.
Finally, the presenter shows that the output can be downloaded as an HTML file (runnable locally without Claude), as a filtered CSV, or as a JSON summary — making the dashboard fully portable and shareable without any ongoing dependency on the AI tool.
Key Insights
- The presenter argues that dataset selection should be driven by the target industry — showing Amazon review analysis to e-commerce companies or stock/tariff data to finance firms like Zerodha makes a portfolio significantly more impactful than using generic datasets.
- The presenter demonstrates a two-pass Claude prompting strategy: first ask Claude to write the best possible prompt for the dashboard task, then copy-paste that generated prompt back into a new Claude session to actually build the dashboard — bypassing the need to know prompt engineering yourself.
- The presenter specifies that the role definition in the prompt must include the industry context (e.g., 'public policy and government consulting') and instructs viewers to swap this portion to match their own target domain such as e-commerce.
- The presenter states that the generated dashboard must include slicers, filters, cross-filtering, drill-downs, tooltips, coordinated charts, KPIs, equity analysis, actionable insights, and strategic recommendations — framing these as non-negotiable elements of a professional, presentation-ready dashboard.
- The presenter highlights that the final Claude output includes raw HTML code, allowing the dashboard to be saved locally and run on any system without needing to open Claude again — and also exportable as a filtered CSV or JSON summary.
Topics
Full transcript available for MurmurCast members
Sign Up to Access