DEV Community: Aditya Raj

LazySort: The Offline AI File Tidier That Organizes Your Mess For You

Aditya Raj — Fri, 23 Oct 2026 12:23:36 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

What I Built
I built LazySort, a background script that watches your messy Downloads folder and automatically categorizes newly saved files into their correct subfolders. It uses a local LLM to read the filename to determine exactly where it should go. It saves me from the tedious chore of manually dragging and dropping files, letting me embrace my inner sloth and be completely lazy about file management.

Demo

Code

badshahaditya57-sketch / LazySort

How I Built It
LazySort is built using Python, the watchdog library to monitor folder changes, and Ollama running a lightweight, open-weight model (llama3.2) locally. When a new file is detected, the Python script sends the filename as a prompt to the local Ollama instance, asking it to classify the file into pre-defined categories. Once the AI responds, Python's shutil moves the file automatically. The entire project centers around fast, local, open-source AI inference to handle the routing logic.

Why Does Open Innovation Matter?
Open innovation matters for LazySort because file organization deals with highly sensitive personal data—like bank statements, health records, or private photos. A closed API would require me to send my private filenames over the internet to a third-party server, which is a massive privacy risk. By using open-weight models running entirely on my own hardware, this automation works completely offline. It guarantees 100% data privacy and costs absolutely nothing in API fees.

The Lazy Payoff
This script saves me about 10-15 minutes of manual sorting every single week. Instead of dreading a cluttered desktop at the end of the week, everything is automatically filed away into "Coursework", "Projects", or "Invoices" folders without me lifting a finger.

LocalList: The Zero-Cost, Private AI Listing Generator for Etsy & eBay Sellers

Aditya Raj — Sun, 18 Oct 2026 15:48:24 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

LocalList is a lightweight, locally hosted web application that generates SEO-optimized product listings for platforms like Etsy and eBay.

It is designed for a very specific side hustle: small business owners, crafters, and vintage resellers who need polished, engaging product descriptions to drive sales, but cannot justify paying a monthly subscription for closed AI tools. LocalList solves the problem of recurring overhead by providing professional-grade AI copywriting entirely for free, running right on the seller's laptop.

Demo

UI Demo Link: LocalList on Streamlit Community Cloud

Code

badshahaditya57-sketch / Local-AI-Listing-Generator-for-Etsy-eBay

🛍️ LocalList: AI Listing Generator for Etsy & eBay

A lightweight, locally hosted web application that generates SEO-optimized, highly engaging product listings for e-commerce platforms like Etsy and eBay.

Built as a submission for the Hacktoberfest Open-Source AI Challenge (Week 2: Side Hustle), this project demonstrates the power of open-weight models by providing small business owners with professional AI copywriting at zero recurring cost and with absolute data privacy.

✨ Features

Zero Subscription Costs: Runs entirely locally using open-weight models, eliminating the need for expensive API tokens.
100% Data Privacy: Your unreleased product ideas, proprietary designs, and business data never leave your machine.
SEO-Optimized Output: Tailors descriptions specifically for Etsy or eBay search algorithms.
Clean UI: Powered by Streamlit for a fast, intuitive, and distraction-free workflow.

🛠️ Tech Stack

Frontend: Python, Streamlit
Local Inference: Ollama
LLM: Llama 3 (8B)
API Integration: Python requests module

🚀 Installation & Setup

Prerequisites

…

View on GitHub

How I Built It

I wanted the technical execution to be clean, fast, and accessible.

Frontend: I used Python and Streamlit to build a simple, intuitive web interface where sellers can input their product name, key features, and target platform.
Open-Source AI Core: The heavy lifting is done using Ollama running the open-weight Llama 3 model directly on my local machine.
Integration: The Streamlit app connects to the local Ollama API endpoint via HTTP requests. It takes the seller's inputs, wraps them in a highly optimized system prompt for e-commerce SEO, and streams the generated listing back to the UI.

Why Does Open Innovation Matter?

For side hustlers, profit margins are razor-thin. Open innovation matters for this build because it fundamentally changes the unit economics of the tool.

If I built this on a closed, general-purpose API, I would have to pass those per-token costs onto the user, likely trapping them in a monthly subscription lock-in. By using an open-weight model running locally, the operational cost is literally zero.

Additionally, open innovation guarantees absolute data privacy. Crafters and designers often have unreleased products or proprietary crafting techniques. Because inference happens locally, their product ideas and drafts never leave their laptop, ensuring their data isn't secretly being used to train someone else's closed model.

My First User

I put this in front of a friend who sells upcycled vintage clothing on Etsy to get real-world validation.

What they said: They loved how fast and straightforward the interface was compared to prompting a generic chatbot.
Would they pay? They explicitly stated they despise monthly subscriptions for seller tools, but would gladly pay a one-time $15 fee to download a packaged desktop application version.
What they asked for next: Their immediate feature request was batch processing—the ability to upload a CSV of 20 unlisted items and have the local AI generate drafts for all of them while they step away to pack orders.

DebugWalk: An AI Walking Mission Generator to Cure Screen Fatigue

Aditya Raj — Tue, 06 Oct 2026 16:59:14 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

I built DebugWalk, a lightweight, client-side web application designed for students and developers who are hitting a wall or suffering from screen fatigue. When you are stuck on a complex coding problem or experiencing digital burnout, you simply type your current task into the app. It then uses AI to generate a highly specific, 15-minute outdoor walking mission metaphorically tied to your problem, forcing you to leave your devices behind and get out into the world.

Demo

Here is the live, deployed web application:
https://badshahaditya57-sketch.github.io/DebugWalk-An-AI-Walking-Mission-Generator-to-Cure-Screen-Fatigue/

Code

badshahaditya57-sketch / DebugWalk-An-AI-Walking-Mission-Generator-to-Cure-Screen-Fatigue

DebugWalk

Ever get so stuck on a bug that you just stare at the screen for an hour? Yeah, me too.

I built DebugWalk for the DEV Community Hacktoberfest "Touch Grass" challenge. It's a super simple web app designed to act as a circuit breaker when you're dealing with screen fatigue or a coding problem that just won't click.

Instead of opening another Stack Overflow tab, you tell the app what you're working on. It then uses AI to generate a quick, 15-minute outdoor walking mission metaphorically tied to your problem. It forces you to step away from the laptop, go outside, and clear your head.

Live Demo

You can try it out right now: [Link to your GitHub Pages URL]

How it's built

I wanted to keep this as lightweight as possible. There is no backend, no database, and no tracking.

Frontend: Plain HTML and Vanilla JavaScript.
Styling: Tailwind…

View on GitHub

How I Built It

I designed this as a zero-backend application using plain HTML, vanilla JavaScript, and Tailwind CSS[cite: 16]. For the intelligence layer, it utilizes prompt engineering via the Gemini API, executing entirely within the browser's local session[cite: 16]. The prompt is strictly engineered to require real-world, screen-free interaction, shifting the AI from a digital assistant to an outdoor catalyst.

Why Does Open Innovation Matter?

Open innovation is crucial for this build because it handles moments of user frustration and burnout. By designing it to run entirely in the local browser session without a backend database, users can freely vent about their tasks without worrying about their data being logged or analyzed by a closed ecosystem[cite: 16]. The open-source repository allows anyone to fork the code, swap in their own preferred open-weight models or local inference endpoints, and host their own free tech-detox tool.

Prize Categories

Use of Partner Technology (optional)

SafeBites: A Zero-Backend AI Meal Planner for My Roommate's Allergies

Aditya Raj — Sat, 03 Oct 2026 18:24:22 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

I built SafeBites, a lightweight, client-side meal planner specifically for my college roommate who has strict dietary restrictions. He frequently struggles to figure out what is safe to cook using the random ingredients left in our fridge. This tool takes our current ingredient list and generates safe, custom recipes tailored exactly to his allergies, solving a daily headache for him.

Demo

Here is the live, deployed web application:
https://badshahaditya57-sketch.github.io/safebites-meal-planner/

Code

https://github.com/badshahaditya57-sketch/safebites-meal-planner/

How I Built It

I kept the stack incredibly simple by building a zero-backend web application using static HTML, vanilla JavaScript, and Tailwind CSS. For the AI component, it utilizes prompt engineering to strictly enforce allergen exclusions via the Gemini API, executing entirely within the browser's local session.

Why Does Open Innovation Matter?

Open innovation was critical for this build because it allowed me to design a tool that requires absolutely no backend infrastructure or databases. By keeping execution strictly client-side, it keeps my roommate's personal dietary and health data completely off external servers that we don't control. It costs us nothing to host, and anyone can fork the open-source repository to run it locally for their own family members.

Prize Categories

Use of Partner Technology (optional)