Data cleaning
Data uploading Section
Evaluation
Training Model(s)
Results
Model Downloadable Pickel File
Learning Guide
Home Page
Model(s) Selection

From "What's ML?" to "I Built That!" – The ML Learning Lab Story

Inspiration

I fell in love with AI after watching a robot beat humans at Jeopardy!—but when I tried to learn machine learning, I hit a wall. Tutorials assumed I knew calculus, GitHub repos looked like alien code, and every error message felt like a personal insult. I thought, "What if there was a tool that made ML as easy as drag-and-drop?"

Enter ML Learning Lab: my Bolt Hackathon project to democratize AI. I wanted to build a playground where anyone—yes, even my art-major roommate—could train models, understand them, and feel like a data scientist.

What it does

✨ No-code ML for everyone!

Upload your dataset (CSV, Excel, or even a messy spreadsheet from your dog's "paw-sonal" budget).
Pick a model (Regression? Classification? I've got you!).
Tweak settings with sliders (because typing random_state=42 shouldn't be a rite of passage).
Train & download your model in minutes, with a Learning Guide explaining every step.
Deploy it straight to a web app (no DevOps black magic required).

How I built it

Tech Stack:

Frontend Technologies

React 18 - Built the entire UI with functional components and hooks (because class components are so 2018).
TypeScript - Saved myself from "undefined is not a function" meltdowns with type safety.
Vite - Blazing-fast builds (because waiting for Webpack feels like watching paint dry).
Tailwind CSS - Made my UI pretty without writing 500 lines of CSS (utility classes FTW!).

UI & Visualization

Lucide React - Fancy icons to make buttons look less lonely.
Recharts - Turned confusing numbers into pretty graphs (even my grandma could understand ROC curves).

PDF Generation

jsPDF + html2canvas - Because everyone wants to print their ML results like a proper scientist.

Development Tools

ESLint - Kept my code clean (no more console.log("debugging desperation") in production).
PostCSS & Autoprefixer - Made sure my app worked even on Internet Explorer (just kidding, nobody supports IE).
React Hooks - Managed state without summoning eldritch horrors (looking at you, Redux).

Key Features

Drag-and-drop ML workflow (upload → clean → train → brag on Twitter).
Interactive data cleaning (fix missing values before they ruin your life).
Algorithm showdown (Logistic Regression vs. Random Forest: FIGHT!).
GridSearch optimization (for when you want to flex that hyperparameter tuning).
Real-time training (with a progress bar so you know it's actually doing something).
One-click PDF reports (for that "I'm a serious data scientist" vibe).
Educational guides (with memes, because why not?).

Deployment

Netlify - Deployed faster than you can say "wait, where's my .env file?"

Challenges I ran into

🚧 The "Oops" Moments:

"Why is my app predicting cats as dogs?" → Forgot to handle categorical data. Fixed with auto-encoding!

What I learned

📚 The Big Lessons:

"No-code" ≠ "No-thinking": UX design is everything for beginners.
ML is scary until it's not: Breaking concepts into emoji-filled steps works ("📉 MSE = How much your model goofs").
Community rocks: Open-sourcing the project brought pull requests from contributors in 20+ countries.