AutoTable: My One-Click Spreadsheet Cleaner Built with Google Gemini

Source: Dev.to

This is a submission for the Built with Google Gemini: Writing Challenge

What I Built with Google Gemini

I built AutoTable, a high‑efficiency, one‑click spreadsheet cleaner designed for knowledge workers who are tired of manual data sanitization. It transforms messy, inconsistent tabular data (CSV, Excel) into validated, ready‑to‑use formats.

The Problem

Knowledge workers spend hours fixing:

• Inconsistent headers
• Mixed data types
• Stray whitespace
• Duplicate rows

Most users don’t want to write complex Excel macros or Python scripts—they just want a fast, trustworthy way to get a clean table.

The Solution

AutoTable handles the heavy lifting with a multi‑stage, idempotent transformation pipeline. Key features:

• Header Normalization – Standardizes messy columns into uniform snake_case.
• Smart Type Detection – Coerces chaotic strings into proper Numbers, Booleans, or Dates.
• Fuzzy Deduplication – Identifies nearly‑identical records using similarity scoring.
• Smart Column Merging – Heuristically combines related fields like “First Name” and “Last Name”.
• Privacy Core – Files are ephemeral and automatically purged, following a “Privacy‑First” architecture.

Gemini’s Role

Google Gemini (via the Antigravity agent) acted as a primary full‑stack pair programmer throughout the development lifecycle. It helped:

• Architect the Next.js App Router structure
• Implement the core cleanData logic
• Design the responsive Tailwind UI
• Write comprehensive test suites using Vitest
• Build complex features such as PayPal subscription integration and Supabase RLS policies

Demo

You can see AutoTable in action here:

Dashboard

What I Learned

Building AutoTable taught me several critical lessons:

• Agentic Workflows – Using AI agents powered by Gemini to generate boilerplate, tests, and core logic dramatically speeds up development.
• State Management in Next.js – Handling complex file‑upload states and real‑time cleaning previews required deep work with React Hook Form and Zod.
• Data Sovereignty – Implementing a “Privacy‑First” design with automated TTL for datasets reinforced secure data handling and Supabase Row‑Level Security.
• Subscription Lifecycles – Integrating the PayPal API for Pro‑tier features involved webhook handling and secure payment verification.

Google Gemini Feedback

What Worked Well

• Code Reasoning – Gemini understood the entire project context across dozens of files and produced functional, pattern‑consistent code.
• Problem Solving – For edge cases like delimiter detection or date parsing, Gemini supplied robust algorithms that saved hours of debugging.
• Architecture Insights – It helped decide how to structure the project for scalability.

Where I Needed More Support

• Visual Fine‑tuning – Achieving the exact “premium” look for glassy UI elements required multiple iterations with Gemini.
• Complex Webhooks – The initial PayPal webhook setup needed additional manual tweaking to handle every possible event state correctly.

Overall, building with Google Gemini felt like having a senior engineer sitting right next to me, ready to tackle any challenge in seconds.