Building AI-powered products from real-world problems | Learning, building, shipping, and improving.

projects

github

productHunt

Surya Kant  • Peerlist

𝐏𝐫𝐨𝐣𝐞𝐜𝐭 𝐎𝐯𝐞𝐫𝐯𝐢𝐞𝐰 NanoErase is a web app built to solve a common digital challenge: efficiently removing watermarks from Google Gemini images. The platform leverages Google API integration to automatically detect and remove watermarks—only for Google Gemini and Nano Banana watermarks—while preserving the original image quality. 𝐏𝐫𝐨𝐣𝐞𝐜𝐭 𝐏𝐮𝐫𝐩𝐨𝐬𝐞 &amp; 𝐕𝐢𝐬𝐢𝐨𝐧: The primary objective of Nano-Erase is to provide a seamless, user-friendly solution for individuals and businesses who need clean, watermark-free images for legitimate purposes such as • Marketing campaigns, • Professional presentations, • Portfolio enhancement, and • Personal use. 𝐓𝐚𝐫𝐠𝐞𝐭 𝐔𝐬𝐞𝐫𝐬: • 𝗣𝗵𝗼𝘁𝗼𝗴𝗿𝗮𝗽𝗵𝗲𝗿𝘀 &amp; 𝗩𝗶𝘀𝘂𝗮𝗹 𝗔𝗿𝘁𝗶𝘀𝘁𝘀 – Remove distracting watermarks from portfolio images • 𝗠𝗮𝗿𝗸𝗲𝘁𝗲𝗿𝘀 &amp; 𝗖𝗼𝗻𝘁𝗲𝗻𝘁 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝘀𝘁𝘀 – Obtain high-quality visuals for promotional materials without watermarks • 𝗖𝗼𝗻𝘁𝗲𝗻𝘁 𝗖𝗿𝗲𝗮𝘁𝗼𝗿𝘀 – Present work without visibility of usage restrictions or watermark artifacts • 𝗕𝘂𝘀𝗶𝗻𝗲𝘀𝘀𝗲𝘀 &amp; 𝗘𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲𝘀 – Streamline workflow for image processing at scale I created this project solely for my personal use, not for selling, but I’m sharing its code for free in my GitHub repo so you can deploy the app on your own hosting platform like Vercel, Netlify, or, Lovable (or any other vibe coding tool)

NanoErase

𝐋𝐢𝐧𝐤𝐞𝐝𝐈𝐧 𝐀𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭 – 𝐅𝐫𝐞𝐞 𝐓𝐞𝐱𝐭 𝐅𝐨𝐫𝐦𝐚𝐭𝐭𝐞𝐫 𝐟𝐨𝐫 𝐄𝐧𝐠𝐚𝐠𝐢𝐧𝐠 𝐏𝐨𝐬𝐭𝐬 📝 𝐏𝐫𝐨𝐛𝐥𝐞𝐦 𝐒𝐭𝐚𝐭𝐞𝐦𝐞𝐧𝐭: • LinkedIn posts often appear plain and monotonous. • Limited formatting options (no bold, italics, styled lists). • Existing tools are either glitchy or paid, making them inaccessible for students and professionals. 💡 𝐒𝐨𝐥𝐮𝐭𝐢𝐨𝐧 • Built a custom text formatting tool using Google AI Studio. • Enables bold, italics, font variations, bullet points, arrows, and structured lists. • Designed to make LinkedIn posts eye-catching and professional. • Completely free to use, shared openly with the community. ⚙️ 𝐃𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭 𝐏𝐫𝐨𝐜𝐞𝐬𝐬 • Explored Google’s vibe coding tools during experimentation. • Identified the gap in LinkedIn content presentation. • Built the prototype in just 1 hour of coding. • Tested with 30+ prompts to validate usability and outcomes. 🎯 𝐈𝐦𝐩𝐚𝐜𝐭 &amp; 𝐎𝐮𝐭𝐜𝐨𝐦𝐞𝐬 • Enhanced readability and engagement on LinkedIn posts. • Supports “Build-in-Public” content pillar (30% distribution). • Achieved 3x engagement multiplier with action-oriented CTAs. • Authentic student-builder voice maintained throughout. 🔑 𝐊𝐞𝐲 𝐃𝐢𝐟𝐟𝐞𝐫𝐞𝐧𝐭𝐢𝐚𝐭𝐨𝐫𝐬 • 100% free, unlike competitors. • Lightweight, glitch-free, and easy to use. • Built by a student innovator, showcasing creativity and problem-solving.<h3>Run Locally</h3>Prerequisites: Node.js<ol><li>Install dependencies: npm install</li><li>Set the GEMINI_API_KEY in <a target="_blank" rel="noopener noreferrer nofollow ugc" class="break-all link-break text-gray-500 hover:underline" href="/Surya17155/LinkedIn-Post-Architect/blob/main/.env.local">.env.local</a> to your Gemini API key</li><li>Run the app: npm run dev</li></ol>Live DemoCheck out the deployed project here: <a target="_blank" rel="nofollow" class="break-all link-break text-gray-500 hover:underline" href="https://wonderful-biscotti-1185db.netlify.app/">https://wonderful-biscotti-1185db.netlify.app/</a>

Linkedin Post Architect

<h3>OmniBridge: Unified AI Provider Gateway</h3>OmniBridge is a high-performance API proxy layer designed to streamline the integration of multiple AI providers. It consolidates fragmented free-tier keys from various AI services, including Gemini, Groq, DeepSeek, GLM, Kimi, and Mistral, into a single, unified OmniKey. This allows developers to point their applications to one OpenAI-compatible endpoint, abstracting away the complexities of managing multiple API keys and provider-specific integrations.<ul><li>Single Endpoint: Utilize one OpenAI-compatible API key to access multiple AI providers, eliminating the need to change your application's code when switching providers.</li><li>Automatic Failover: Seamlessly handles provider outages, rate limits, or quota exhaustion by automatically routing requests to the next available provider.</li><li>Smart Routing: Distributes general queries across providers and intelligently routes task-specific requests (e.g., code generation to DeepSeek, image processing to vision-capable models).</li><li>Extensive Provider Support: Integrates with over 16 AI providers, including OpenAI, Anthropic, Google (Gemini), Groq, Together, and more.</li><li>Dashboard: Provides a centralized interface for managing API keys, configuring provider rotation, and monitoring usage statistics.</li><li>Proxy API: Enables the use of any provider's models through OmniBridge's proxy, with automatic key rotation for continuous access.</li></ul>The tech stack includes React for the frontend, TypeScript for type safety, Node.js for the backend, and deployment on Vercel. It supports databases like Turso (libSQL) or local SQLite, and employs AES-256-GCM for secure storage of API keys.Getting started involves installing dependencies with npm, setting up environment variables, and running the development server with npm run dev. The architecture features a client application interacting with the OmniBridge API proxy, which then routes requests to various AI providers.

Omnibridge

𝐕𝐢𝐛𝐞𝐎𝐧 — A modern, production-ready website built in 24 hours using 5 AI tools (Antigravity, GLM-5V-Turbo, MotionSites.ai, Dazl.dev, Lovable). 𝐖𝐡𝐲 𝐈 𝐛𝐮𝐢𝐥𝐭 𝐭𝐡𝐢𝐬: A college Digital Marketing assignment became an opportunity to learn real product building. Instead of a throwaway submission, I created a reusable frontend template. 𝐖𝐡𝐚𝐭 𝐢𝐭 𝐭𝐚𝐮𝐠𝐡𝐭 𝐦𝐞: • How to use agentic AI tools to design + deploy without coding • Real workflow: design → component building → integration → deployment • 𝐖𝐡𝐚𝐭 𝐢𝐭 𝐢𝐬: A clean, modern landing page inspired by webild.io—ready to be transformed into a micro-SaaS product. The structure stays. Only the details change. 𝐊𝐞𝐲 𝐥𝐞𝐚𝐫𝐧𝐢𝐧𝐠: Use assignments as real-world project opportunities. Learn AI tools. Build products. Ship them.

VibeOn

a Custom GPT that writes realistic image JSON prompts for Nano Banana Pro. • 𝗧𝗵𝗲 𝗣𝗿𝗼𝗯𝗹𝗲𝗺: Most AI image prompts are too generic. Result? Mediocre, unrealistic outputs. • 𝗧𝗵𝗲 𝗦𝗼𝗹𝘂𝘁𝗶𝗼𝗻: A trained GPT that creates detailed JSON prompts specifically optimised for Nano Banana Pro. It covers every minor detail of an image, like a Professional Photographer, like lightings, camera angle, aperture, mm lens, realistic skin texture details, trained with 7+ detailed PDF guides to create realistic image prompts.

Prompt Banana

This is an MVP/Prototype of a concept by which Lawyers and Advocates can manage their clients and cases, They can create their own organisation in it and can add team members in it. The AI will summaries their case details and set the dates on their Google calender automatically, they can ask questions to AI about their case, summaries, key points, etc.. Today almost every Lawyer and Advocate recording their entries in books, I created this project under Lexicorn to make this process digitalized. The physical books can be stolen, can be damaged, but this digital version of their books cannot be destroyed and they have an AI agent inside it they can have chat with it about their case and it can automatically set their court timings based on the cases and manual prompts.

LegalCase Pro

This chatbot is designed for students who are searching for colleges and have come across IILM University, Greater Noida. They may only know the name of the college and want to learn more about it. Here are some options they have: - Read the information on the website (which some may find boring). - Call customer support. However, many students tend to search for this type of information at night, and there’s typically no one available to answer their calls then. This is where an AI chatbot can be particularly useful. Students can chat with it anytime and anywhere. Universities and colleges advertise on platforms like Instagram, and students can even interact with this chatbot through Instagram DMs. If they prefer not to type, the chatbot is also available in voice mode, powered by Vapi AI.