DEV Community: Titus-Dsouza

👻 Ghost in the Machine: A Spooky Gemini-Powered AI Experience

Titus-Dsouza — Wed, 28 Oct 2026 13:28:33 +0000

What I Built

Ghost in the Machine is a Halloween-themed interactive website that brings together spooky visuals and artificial intelligence.

Inspired by the idea of a haunted computer, the project features a playful dark interface, floating ghost stickers, Halloween decorations, an interactive haunted terminal, and AI-generated spooky stories.

Rather than creating a serious horror experience, I wanted to make something friendly, fun, and interactive while exploring how AI can be integrated into a web application.

Demo

The project is currently under development.

A live demo will be added once deployment is complete.

Code

👻 GitHub Repository: https://github.com/Titus-Dsouza/Ghost-in-the-Machine

How I Built It

The project uses:

HTML5 for the website structure.
CSS3 for the dark Halloween theme, animations, and responsive design.
JavaScript for interactive elements and communication with the backend.
Python and Flask for backend API routes.
Google Gemini API for chatbot responses and spooky story generation.

The intended architecture connects the frontend to a Flask backend, which sends requests to Gemini and returns the generated responses to the website.

Why Does Open Innovation Matter?

Open innovation makes it easier for developers to experiment, learn from existing tools, share their work, and build on one another's ideas.

This project is an opportunity for me to explore AI integration, frontend development, backend APIs, and open-source project development. I also want to keep the code organized and accessible so others can understand the implementation and contribute improvements.

Although this project uses the Gemini API, I recognize that an API-based model is not itself necessarily open-source. The open-source aspect of this project is the code and the opportunity for others to inspect, learn from, and extend it.

My Agent Session

I used an AI coding agent to help develop and refine the project. I may add a session recording here if I publish one through DevRelay.

The Spirit of It

What if the computer you were coding on had a ghost living inside it?

That idea inspired Ghost in the Machine: a place where forgotten code, spooky stories, and AI meet in a playful Halloween experience.

The goal is to make experimenting with AI feel less intimidating and more creative—one ghostly message at a time. 👻🎃

Prize Categories

This submission is for the Hacktoberfest Open-Source AI Challenge Week 4. Additional partner prize categories will be listed if applicable to the completed project.

Thanks for checking out my project! 👻

StationerySort: Let Local AI Handle the Boring Inventory Updates

Titus-Dsouza — Sat, 24 Oct 2026 09:58:12 +0000

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

What I Built

I built StationerySort, a lightweight AI-powered stationery inventory manager designed to make everyday organization easier for students.

Keeping track of pens, notebooks, sticky notes, and other supplies can become repetitive. Instead of manually updating quantities every time something is purchased or used, StationerySort lets users describe inventory changes in plain English.

For example, you can type "I bought 10 new blue pens" or "Used 2 sticky note pads." The AI generates a proposed inventory update, which the user can review and confirm before it is saved.

The application also tracks stock levels, highlights items that need restocking, and provides search and category filters.

Demo

Images are provided and git hub repo:https://github.com/Titus-Dsouza/StationerySort

Code

GitHub Repository: https://github.com/Titus-Dsouza/StationerySort

How I Built It

I built StationerySort using:

Python and Flask for the backend.
SQLite for storing stationery inventory.
HTML, CSS, and vanilla JavaScript for the frontend.
Gemma through Ollama for local, open-weight AI inference.

The AI assistant interprets natural-language inventory updates and generates structured proposals. A fallback rule-based parser allows the application to remain usable when Ollama is unavailable.

I also included a human-confirmation step so that AI-generated changes are reviewed before being applied to the database.

Why Does Open Innovation Matter?

For this project, open-weight AI is more than an optional feature. It is part of the actual workflow.

Using Gemma through Ollama gives users the flexibility to run a model locally rather than depend entirely on a proprietary cloud AI API. It also avoids per-request AI API charges when inference is performed locally and allows users to experiment with different compatible models.

Local inference helps keep inventory information on the user's own machine, while the fallback parser makes the application more resilient when the model runtime is unavailable.

For me, open innovation means being able to experiment, learn, and build useful AI applications without making a paid cloud service a mandatory dependency.

The Lazy Payoff

The idea behind StationerySort is simple: spend less time maintaining inventory and more time getting things done.

A short natural-language instruction can replace several manual steps, while automatic low-stock alerts help users identify what needs restocking.

It is a small project, but it explores a practical way to make everyday tasks a little less tedious.

Prize Categories

Open-source AI / open-weight models, as applicable to the challenge's eligibility criteria.

Eden Homes: Bringing a Local Homestay Business Online with Open-Source AI

Titus-Dsouza — Sun, 18 Oct 2026 17:58:54 +0000

![ ](https://dev-to-uploads.s3.us-east-2.amazonaws.com/uploads/articles/ev2a4lywb46x96mfbmdi.png)
![ ](https://dev-to-uploads.s3.us-east-2.amazonaws.com/uploads/articles/dbm6t65dnfk3qv2f2n9s.png)
![ ](https://dev-to-uploads.s3.us-east-2.amazonaws.com/uploads/articles/59173z3xx4mg910v90ho.png)

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

What I Built

Eden Homes is a website project for a homestay business in Honnavar, Karnataka, India.

Small hospitality businesses often rely on phone calls and messaging to handle enquiries. I wanted to build a straightforward online experience where travellers can explore accommodation options and contact the owner without navigating a complicated booking platform.

Eden Homes focuses on two accommodation options: 1 BHK and 2 BHK.

The website is designed to help visitors:

Explore property options through image-led listings.
View property photos in a gallery.
Open a booking enquiry form for their selected property.
Enter their name, phone number, and email address.
Contact the owner directly by phone if they need more information.

The intended customers are families and tourists visiting Honnavar. The business could benefit from more enquiries and less manual effort managing potential bookings.

My goal is to keep the experience simple, accessible, and suitable for a small local business rather than building another complicated booking platform.

Code

GitHub repository: https://github.com/Titus-Dsouza/Eden-Homes

The project currently uses HTML, CSS, and vanilla JavaScript for the frontend, with Supabase being integrated for backend services.

I'm developing the codebase in phases so that the interface, database, authentication, and future AI features can be built and tested incrementally.

How I Built It

I started with a simple frontend and developed the property listings and booking enquiry interface.

Current technologies:

HTML5 — page structure and forms.
CSS3 — responsive styling and layout.
Vanilla JavaScript — property selection, gallery interactions, and booking form behaviour.
Supabase — database and authentication services, with an Edge Function intended to process booking enquiries.

I'm approaching the project phase by phase rather than trying to build everything at once. This helps me understand how the frontend, backend services, and database fit together.

I'm also exploring how open-weight AI models could support an Eden Homes AI concierge in the future. The goal is to help visitors understand accommodation options and get answers to common questions.

The AI concierge is a planned extension, not a completed feature. I want to integrate it in a way that is useful for the business and practical to maintain.

Why Does Open Innovation Matter?

For a small business, the cost of building and maintaining software matters.

Open innovation makes it possible to experiment with open-weight AI models, inspect how components work, and choose infrastructure according to the project's needs. It also offers the flexibility to explore locally running models instead of depending entirely on a paid, closed AI API.

For Eden Homes, this could eventually mean a more affordable AI concierge that answers frequently asked questions, explains property options, and helps visitors prepare booking enquiries.

Open innovation also makes learning more accessible. As a student developer, I can experiment with the technology, understand its limitations, and build toward a practical use case without making expensive AI services a requirement from day one.

My aim is not to add AI just because it is trending. It is to explore whether AI can solve a real problem for a small hospitality business.

Prize Categories

Challenge: Hacktoberfest Open-Source AI Challenge Week 2 — Side Hustle.

I'm focusing on building a practical small-business project and exploring how open-source AI can become a useful part of it.

I'm building Eden Homes incrementally, learning along the way, and aiming to turn a simple local homestay website into a practical tool for a real small business.

Feedback, suggestions, and contributions are welcome!

TouchGrass 🌿 — Turning Screen Time into Outdoor Time

Titus-Dsouza — Mon, 12 Oct 2026 06:32:31 +0000

Submission for Hacktoberfest Open-Source AI Challenge — Week 1: Touch Grass

What I Built

TouchGrass is a web app concept designed to help people spend less time scrolling and more time exploring the world around them.

The idea is simple: instead of spending more time deciding what to do, users should be able to get a practical outdoor activity plan and go try it.

TouchGrass aims to help users discover short outdoor activities based on factors such as their available time, energy, mood, and interests. These could include taking a walk, exploring a nearby area, spending time in nature, or trying a simple outdoor challenge.

The core philosophy is:

Use AI → Get a plan → Leave the screen.

The current implementation includes a rule-based fallback. Real inference from an AI model is not yet verified, so the project is still a work in progress.

Demo

Repository: https://github.com/Titus-Dsouza/Hacktober_fest_week_2
Live demo: A public deployment URL still needs to be confirmed.

The Vercel link currently available is the project dashboard: https://vercel.com/titus-dsouza/hacktober_fest_week_2

How I Built It

The project uses the following planned technology stack:

Frontend: HTML, CSS, and JavaScript
Backend: Python with Django and Django REST Framework
Authentication: Supabase Auth
AI integration: Ollama with qwen2.5:3b planned for local, open-weight inference
Fallback: Rule-based responses when model inference is unavailable

The current working fallback helps establish the basic experience while AI integration remains a work in progress.

Why I Chose This Idea

Technology is useful, but it can also keep us indoors and glued to our screens for longer than we intend.

I wanted to explore a project that uses technology to encourage people to step away from it. Rather than building another chatbot that encourages endless conversation, TouchGrass is intended to help users find an activity and then get off the screen.

My goal is to develop the project further by making outdoor suggestions more useful, personal, and accessible.

Challenges and What I Learned

One of the main challenges is building a useful experience before the AI integration is fully verified. A rule-based fallback provides a foundation, but connecting and testing a real open-weight model is an important next step.

This project is also helping me learn how frontend interfaces, backend APIs, authentication, and AI components fit together in a web application.

What's Next?

Verify real inference with the open-weight model.
Test the complete user journey in a browser.
Improve personalization and the quality of outdoor missions.
Publish a working public demo.
Add screenshots and improve setup documentation so others can try the project.

Prize Category

Open Innovation

Final Thought

TouchGrass is a work in progress, but the mission behind it is straightforward: use technology to help people experience more of the world beyond their screens.

Build something useful. Get outside. Touch grass. 🌿

hf26challenge #Hacktoberfest #OpenSource #AI #Django #TouchGrass

Starting my developer journey 🚀

Titus-Dsouza — Mon, 05 Oct 2026 07:30:31 +0000

Excited to learn, build, try new things, and complete challenges along the way.

DevChallenge #BuildInPublic #Coding