EvidencePlot AI

Inspiration

I am passionate about building technology that makes life easier for people. With my knowledge of data analysis and research methodology, I understand some of the challenges researchers face when working with data.

I wanted to build something practical that could help researchers get reliable statistical results and professional scientific figures without needing to write complex code.

Before settling on this idea, I conducted market research to identify a product that solves a real problem and has the potential to become a sustainable business. That process led me to build EvidencePlot AI.

What it does

EvidencePlot AI helps researchers and students turn raw CSV data into scientific figures and statistical results they can verify.

Users can upload their data, get AI-assisted recommendations for analysis methods, review important assumptions, run statistical tests, and customize their figures.

The app also creates a Figure Evidence Passport that connects the results to the dataset and calculation details. Instead of relying on AI to guess statistical answers, it uses Python and SciPy to perform the actual calculations.

Privacy is a priority. Raw research data stays in the user's browser, while only limited, sanitized metadata is sent to the AI service when AI assistance is enabled.

How I built it

I approached this project as an architect. My role was to define the problem, understand the requirements, guide the development process, and ensure the final product followed the intended design and technical approach.

I used Google Antigravity, an agentic IDE, and the provided SKILL.md file to guide the AI development agent through the project instructions step by step. Rather than building every component manually, I used this workflow to turn the project requirements into a working application.

The application uses React and TypeScript for the interface, Pyodide with Python and SciPy for statistical calculations in the browser, and Cloudflare Workers AI for recommending analysis methods.

I also worked through the requirements for data privacy, statistical accuracy, claim verification, and automated testing. This experience showed me how a clear plan, the right tools, and an AI agent can help turn an idea into a working product.

Challenges I faced

One of my biggest challenges was the cost of development. I live in Nigeria, and building with advanced AI tools required subscriptions paid for in US dollars. These costs, together with other expenses such as electricity, made it challenging to keep developing the project.

Another challenge was ensuring that the application did more than simply generate convincing answers. Scientific research requires accuracy and trust. I needed a system where actual statistical calculations, rather than AI guesses, determined the numerical results.

I also had to address data privacy and ensure that heavy calculations did not freeze the user interface. These challenges pushed me to think carefully about the architecture and how the different parts of the application should work together.

What I learned

I learned that building with AI is not just about writing prompts. It requires understanding the problem, designing the system, giving clear instructions, following a structured workflow, testing the results, and checking that the final product works as intended.

I also learned that you do not have to do everything manually to build something valuable. With the right technical understanding, planning, and AI tools, you can guide an agent to help you build a complex application.

What's next

My goal is to grow EvidencePlot AI into a useful product that helps researchers and students around the world. I want to improve it, reach more users, and build a sustainable business around it.

Winning this competition would give me an opportunity to access resources that can help me develop and market the product further.

Beyond this project, I want to use what I learn and earn to help more people, especially those who may not have easy access to expensive technology tools. I want to teach people how to build with AI, learn coding, and create solutions that improve their lives.

I believe this new way of building software can create opportunities for people who have ideas but lack the resources to bring them to life. With the right support and resources, I believe I can do more, teach more people, and help others use technology to create better opportunities for themselves.

EvidencePlot AI is one step toward that vision: using AI not just to make work easier, but to empower people to build things that matter.

Built With

• artificial-intelligence
• cloudflare-workers
• cloudflare-workers-ai
• csv-processing
• data
• data-visualization
• llama-3.3
• pyodide
• python
• react
• scientific-research
• scipy
• statistical-analysis
• tailwind-css
• typescript
• vite
• vitest
• web-crypto-api
• web-workers
• zod