DEV Community: Soumasish Dasgupta

BrandSpark: AI-Powered Logo Generator with Google AI Studio

Soumasish Dasgupta — Fri, 04 Jul 2025 14:52:25 +0000

This post is my submission for DEV Education Track: Build Apps with Google AI Studio.

What I Built

I created BrandSpark, a web app that leverages Google Gemini and Imagen models (via Google AI Studio) to generate complete branding kits in seconds.

Key prompt used:
- “Create an app which uses Al to generate unique logos and branding descriptions for businesses, projects, and teams. Users provide details and receive a custom logo and tagline created by Gemini.”
Additional features:
- Choice of logo styles (Modern, Retro, Geometric, etc.)
- Predefined color palettes (Vibrant, Earthy, Monochromatic, etc.)
- Downloadable PNG assets and persistent local state

Demo

Here’s a quick look at BrandSpark in action:

Also you can check out the website at: BrandSpark

My Experience

Working through the Build Apps with Google AI Studio track taught me:

Seamless API Integration:

How to structure requests to Google Gemini for text‑based prompt engineering.
Chaining responses into Gemini model calls for high‑quality image outputs.

Prompt Engineering Nuance:

Small tweaks in phrasing dramatically affected logo composition and color balance.
Iterating on prompts led to much crisper, on‑brand visuals.

State Management & UX Considerations:

Saving user inputs/results in localStorage made the experience frictionless.
Implementing spinners and error messages ensured smooth, transparent feedback.

What surprised me most was how quickly the AI iterated on dozens of design variations—the combination of Google’s large‑language and image models felt truly magical. This project sharpened my full‑stack skills (React, Vite, Tailwind) while deepening my understanding of modern generative‑AI workflows.

ToneShift: Transforming Communication with AI

Soumasish Dasgupta — Sun, 19 Jan 2025 18:38:00 +0000

This is a submission for the GitHub Copilot Challenge : Transitions and Transformations

What I Built

ToneShift is a cutting-edge tool that leverages AI to enhance text communication by transforming its tone and sentiment. It enables users to seamlessly adapt their messages to fit a range of contexts, from formal and professional to empathetic and motivational.
It helps users maintain clarity and emotional intelligence in their written interactions.

Demo

You can check out the website here: ToneShift
Watch ToneShift in action here

Repo

The full source code is publicly available on GitHub:

Soumasish2005 / toneshift

ToneShift is an AI-powered tool designed to transform the tone of your text to suit various contexts. It ensures your communication is clear, empathetic, and effective.

ToneShift: AI-Powered Tone Transformation

Built as a part of the GitHub Copilot Challenge , ToneShift is an AI-powered tool designed to transform the tone of your text to suit various contexts. Whether you need to adjust your message for professional, personal, or cultural purposes, ToneShift ensures your communication is clear, empathetic, and effective.

Key Features

Tone Transformation Options
- Predefined tones: Professional, Empathetic, Motivational, Casual, Formal, Neutral.
- Automatic detection of the tone of the text via AI.
- Customizable tones for fine-tuning emotional intensity.
Real-Time Integration
- Works as a web application for transforming text.
- Provides real-time suggestions and edits as you type.
Personalized Recommendations
- Adapts to your preferences with machine learning over time.
- Saves and applies user-defined tone profiles.

Use Cases

Professional Communication: Convert casual messages into formal emails or transform critical feedback into constructive input.
Conflict Resolution: Rewrite emotionally charged messages to ensure diplomacy and understanding.
Customer Support…

View on GitHub

Copilot Experience

Throughout the 4.5 hours of development, GitHub Copilot helped me a lot in :

Prompts: Used Copilot to generate boilerplate code for React components and Node.js API routes.
Edits: Enabled rapid iteration by suggesting improvements to existing code.
Chat: Consulted Copilot Chat for debugging and resolving specific issues with NLP model integration.
Autocomplete: Accelerated development by leveraging intelligent autocompletion for repetitive tasks.
Model Switcher: Experimented with different model configurations to optimize tone transformation accuracy.
generating unit tests: Copilot also helped me a lot in generating test cases for testing the backend api via postman.

GitHub Models

Yes, GitHub Models played a significant role in prototyping the LLM capabilities for ToneShift. Here’s how they were utilized:

Sentiment Analysis: Prototyped and fine-tuned LLM-based sentiment detection using Models like GPT 4o and Llama.
Tone Transformation: Leveraged GitHub Models to generate tone-altered versions of input text with high accuracy.
Structuring prompt: Made the prompt more concise and refined for improved accuracy of output. Gpt 4o, claude 3.5 sonnet were used in this process.

Conclusion

ToneShift maintains the theme of Transitions and Transformations by re-imagining how communication evolves to meet diverse needs. It transforms written text to align with desired tones, empowering users to navigate shifts in context, from casual to formal or critical to empathetic.

The potential impact of this project is vast: it can enhance professional relationships, improve conflict resolution, and foster greater understanding across cultural and emotional divides. By making communication more adaptable and empathetic, ToneShift contributes to building connections that are both meaningful and effective.

This experience has taught me the importance of ethical AI in facilitating human interactions.

Some future expansions, such as voice integration and creation of a browser extension along with integration to platforms like slack, Gmail, Outlook etc., will further broaden its applicability and societal impact.

Start transforming the way you communicate today with ToneShift!

Thank you for considering my submission. Feel free to write anything in the comments...❤️❤️

EcoStarter: Empowering Fresh Starts for a Sustainable Future

Soumasish Dasgupta — Tue, 14 Jan 2025 16:38:52 +0000

This is a submission for the GitHub Copilot Challenge : Fresh Starts and/or New Beginnings

What I Built

EcoStarter is a web app that empowers individuals to begin their journey towards a sustainable lifestyle. It provides tools for lifestyle analysis, personalized eco-friendly challenges, a carbon footprint tracker, and a community space to share ideas and progress. The app focuses on helping users reduce their environmental impact through actionable steps and measurable progress.

Key Features:

Lifestyle Analysis: Understand current habits and receive personalized AI-driven recommendations to adopt eco-friendly practices.

Eco Challenges Dashboard: Create and complete challenges which will help you progress towards a more sustainable lifestyle.

Carbon Footprint Tracker: Track the environmental impact of your lifestyle over time using dedicated metrics.

Community Space: Connect with others, share your journey, and witness people's journey towards a sustainable journey. (See demo)
Resource Finder: Enhance your knowledge by exploring a curated collection of blogs, books and other resources on adopting sustainable lifestyles.

Demo

You can check out the deployed app here: EcoStarter Live Demo

Repo

The source code for the project is available on GitHub: EcoStarter

Soumasish2005 / EcoStarter

EcoStarter

EcoStarter is a web application designed to help individuals adopt a more sustainable lifestyle. It provides personalized eco-friendly recommendations, tracks carbon footprints, and connects users with a community of like-minded individuals.

Features

Personalized Plan: Get a customized eco-friendly lifestyle plan based on your habits and goals.
Daily Challenges: Complete fun and meaningful eco-challenges to build sustainable habits.
Community Support: Join a community of eco-conscious individuals sharing tips and experiences.
Impact Tracking: Measure and visualize your environmental impact over time.

Tech Stack

Frontend: React, TypeScript, Tailwind CSS
Backend: Firebase (Firestore, Authentication)
APIs: Cloudinary for image uploads, Google Generative AI for lifestyle analysis

Getting Started

Prerequisites

Node.js
npm or yarn

Installation

Clone the repository:

git clone https://github.com/yourusername/EcoStarter.git
cd EcoStarter

Install dependencies:
```
npm install
```
Set up environment variables Create a .env file in the root directory and add your Firebase and Cloudinary configuration:
```
VITE_FIREBASE_API_KEY=
```
…

View on GitHub

Copilot Experience

Throughout the 8 hours of developing EcoStarter, GitHub Copilot helped me a lot. I used the various features Copilot has to offer so that I could make the project better.

Code Auto-completion : I used Copilot's code autocompletion to generate boilerplate for the components. This feature also helped me complete code faster as it automatically completed the code as I kept typing, thereby making development faster.
In-line chat: I used the inline chat feature to fix errors as well as generate Tailwind CSS styling in some areas.

Generating Tests: Copilot also helped me generate sample resources for testing my code as well as displaying it on the website. This helped me a lot as I didn't have to find resources manually and was able to save time.
Bug Fixing & debugging: Copilot suggested fixes for some common coding errors on the go, which made debugging code very easy. I also took Copilot's help to understand and debug various errors which I got in the browser console. I would tell Copilot to explain the error and Copilot gave a detailed explanation of why the error was working along with possible solutions which I could just copy and paste from the chat tab.
Copilot Chat: Copilot chat helped me code the authStore and userStore hooks with the help of Zustand.

Overall, using GitHub Copilot was a great experience, as it helped me code faster and with fewer errors.

GitHub Models

I used two models in this project:

GPT 4o
o1(Preview)

These models helped me phrase the prompts properly so that I could get a better response from Google generative AI while trying to calculate carbon emissions as well as to suggest sustainable lifestyle to users.
I also checked out other models like Claude 3.5 sonnet (Preview) but as it was giving large code blocks with redundant code in some places, I omitted using it. Overall, this project introduced me to GitHub models, and I'm really amazed to see how good these models are. I look forward to using them with this project in future as well.

Conclusion

Building EcoStarter was a rewarding experience for me, as it taught me how I can combine AI with sustainable practices and user engagement to create an impactful solution.

The primary purpose of building this web-app is to enable users to take their first steps towards a more eco-friendly life. This app empowers users to break old habits and make a "fresh start" with a focus on sustainability.
This app also plans to reduce the carbon-footprints of users and making them more conscious about the environment. The features like the carbon footprint tracker, personalized challenges, and resource finder actively promote responsible consumption and climate action.

Future Plans:

Even though this contest was of 24 hours, I would still like to continue improving this app by adding the following features:

Dedicated section for recycling: I would like to add features for locating nearby recycling centers, and eco-friendly vendors.
Improvements to the community section: I would like to add some more features to the community section
Gamification of the website: I would like to add experience points(XPs) to each activity which would also be used to redeem some rewards and stuff. In this way, users will feel more engaged.
Adding more challenges: I will also add more personalized challenges(daily, weekly, monthly) in the challenges section.

Thanks for giving me an opportunity to work on this beautiful project as part of this hackathon. I really hope people find this website useful. Do share your thoughts in the Comments section💖💖.....