I Built an AI That Organizes My Files Because I Was Too Lazy To Do It 🛋️

Gayathri Ande — Wed, 21 Oct 2026 19:07:02 +0000

This is my submission for the Hacktoberfest 2026 Week 3 challenge ("Slow and Steady").

My Downloads folder was a graveyard

My Downloads folder wasn't a folder anymore. It was a graveyard.

Every file followed the same cycle: download it, mean to rename it, mean to move it, postpone it forever. scan0001.txt. resume_final_v3. untitled. I never decided to ignore them. Sorting them was just too boring to start.

The idea

LazyDesk is a local-first file organizer. You drop files in. A model running on your own machine reads them, decides what they are, suggests a clean filename, and moves them into folders once you approve.

Drop it. Forget it.

![Messy folder before LazyDesk](

)

What LazyDesk does

For each file, LazyDesk:

Extracts text (plain text, PDF, DOCX; images are judged by name and type only).
Asks the model for structured JSON: category, suggested_filename, summary, confidence, reason.
Shows the proposal before anything happens, for example scan0001.txt becomes Documents/Invoice_1042.txt.
Moves files only after you approve.

![LazyDesk results screen](
![ ](

Architecture

React + TypeScript UI
        ↓
FastAPI backend
        ↓
File extractor (text / PDF / DOCX)
        ↓
Ollama → [Gemma model name]
        ↓
JSON validation + confidence check
        ↓
Your approval
        ↓
Organizer → workspace/organized/<Category>/

Why open-weight, local AI?

Folders like this hold resumes, certificates and private notes. I didn't want to upload them to a third-party AI service just to rename them.

I used [Gemma model name] through Ollama on [my machine: CPU/GPU, RAM]. The tradeoffs are real: local inference needs capable hardware and can be slower than a hosted API. It isn't absolute privacy either, because the files still live on my disk. What I get is control, the ability to swap models, and no per-request API cost.

Building it in one day

I only had one day, so I locked the scope: no authentication, no database, no RAG, no OCR. A few decisions mattered most.

Fixed categories. Documents, Certificates, Resumes, Projects, Notes, Images, Other. A fixed list makes the model's output easy to validate. [What did you actually see? Did it help?]
The model never touches paths. It returns a category and a filename. My code validates the category, sanitizes the name, and picks the destination inside one workspace folder.
Safe by default. Duplicate names get _2, nothing is overwritten, and low-confidence or invalid results land in "Needs Review", unapproved.

The lazy payoff

I gave LazyDesk [N] files. It prepared [X] automatically and flagged [Y] for review. Analysis took [T] seconds on [machine]. Sorting the same files by hand took me [M]. Of the [N] classifications, [K] were correct.

![Organized folders after LazyDesk](

)

Demo

Limitations

Quality depends on the model; small models misclassify.
No OCR, so screenshots are judged only by filename and type.
Tested on [N] files. This is a demo, not a benchmark.
Local inference needs enough hardware to run the model.

Future work

OCR for screenshots, duplicate detection, a watched inbox folder, and undo.

Source code

gayeeA / lazydesk

LazyDesk 🛋️ Drop it. Forget it.

A local-first AI file organizer. Drop messy files in; an open-weight model running on your machine classifies them, proposes clean filenames, and, after you approve, moves them into category folders.

Problem

Downloads and Desktop fill up with resumes, certificates, notes and scans. Renaming and sorting them is boring, so people postpone it.

Solution

LazyDesk extracts text from each file, asks a local model (Gemma via Ollama by default) for structured JSON, shows the proposed actions, and organizes only what you approve.

Why open-weight / local AI?

Files can be private. The model runs on your machine through Ollama, so LazyDesk makes no cloud AI calls. Tradeoffs: it needs capable hardware and can be slower than a hosted API. This removes a third-party AI service from the loop; it does not make the whole workflow "absolutely private".

How it works

React UI → FastAPI →…

View on GitHub

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

😂 I Built a Resume Generator Based on Mood (Worst Career Tool Ever)

Gayathri Ande — Sat, 11 Apr 2026 18:26:02 +0000

This is a submission for the DEV April Fools Challenge

😂 I Built a Resume Generator Based on Mood (Worst Career Tool Ever)

😂 What I Built

Ever wondered what your resume would look like if your emotions took over?

I built a Mood-Based Resume Generator that creates completely useless (but hilarious) resumes based on your mood.

Instead of helping you get a job, this tool will probably make recruiters block you 😄

🎮 Demo

👉 Live App: https://gayeea.github.io/Mood_based_resume/

👉 GitHub Repo: https://github.com/gayeeA/Mood_based_resume/

💻 Code

The project is built using simple frontend technologies. Here's the core idea:

User enters name & role
Selects a mood
JavaScript generates a chaotic resume

Example logic:

if (mood === "angry") {
  return "I dominated every project and crushed deadlines. Hire me or regret it.";
}

⚙️ How I Built It

HTML for structure
CSS for styling
JavaScript for generating mood-based content

No backend. No database. No purpose. Just vibes.

😂 Example Outputs

😡 Angry Mode

"I dominated every project and crushed deadlines. Hire me or regret it."

😭 Sad Mode

"I tried my best… sometimes things worked… sometimes not…"

😎 Overconfident Mode

"I am the greatest developer alive. My code is flawless."

🤖 Robot Mode

"Human executed assigned tasks. Efficiency: HIGH."

🤡 Why It's Useless

Because no recruiter would ever accept a resume like this.

But it does answer an important question:
👉 What if your feelings wrote your resume?

🏆 Prize Category

Community Favorite

Because this project is:

Funny
Relatable
Slightly painful

🚀 Final Thoughts

This was built purely for fun and chaos.

Sometimes, the best way to learn is to stop being serious and just build something ridiculous.

Mission: accomplished 😄

DEV Community: Gayathri Ande

I Built an AI That Organizes My Files Because I Was Too Lazy To Do It 🛋️

My Downloads folder was a graveyard

The idea

What LazyDesk does

Architecture

Why open-weight, local AI?

Building it in one day

The lazy payoff

Demo

Limitations

Future work

Source code

gayeeA / lazydesk

LazyDesk 🛋️ Drop it. Forget it.

Problem

Solution

Why open-weight / local AI?

How it works

😂 I Built a Resume Generator Based on Mood (Worst Career Tool Ever)

😂 I Built a Resume Generator Based on Mood (Worst Career Tool Ever)

😂 What I Built

🎮 Demo

👉 GitHub Repo: https://github.com/gayeeA/Mood_based_resume/

💻 Code

⚙️ How I Built It

😂 Example Outputs

🤡 Why It's Useless

🏆 Prize Category

🚀 Final Thoughts