Open-source workplace AI / enterprise context layer for permissioned search, RAG, and agentic workflows across 50+ connectors. Contributing as an SDE — helping extend the governed context layer for explainable enterprise search, MCP, and self-hosted AI automation.
Open-source voice AI platform: real-time STT → LLM (tool-calling, RAG, human transfer) → TTS over SIP telephony or browser testing. Contributing as an SDE on the voice pipeline, gateway, and self-hosted Docker stack.
Open-source, self-hostable voice AI platform (Vapi/Retell alternative) with a visual workflow builder, BYOK LLM/STT/TTS, telephony, and MCP. Contributing as an SDE on production voice-agent orchestration and self-hosted deployments.
A multi-tenant dealership and automotive management application for showroom visitor management with WhatsApp integration, digital & field enquiries, delivery updates, vehicle models, lead sources, and templates. Includes RBAC, organization-level feature toggles, lead/CRM workflows, and has been deployed to 5+ showrooms with 5000+ visitor entries.
RukiAI is an AI-driven personal finance tracker designed to help users log expenses, set budgets, and receive smart, personalized financial advice. Built with Node.js, MongoDB, and EJS, it integrates Cohere's AI API to deliver real-time insights while also featuring a custom local AI model to demonstrate advanced AI capabilities for recruiters. The project showcases both cloud-based AI integration and self-hosted model development in one application.
A full-stack MERN application for managing WhatsApp marketing campaigns with role-based access for admins and resellers. Features include client onboarding, financial transaction tracking with credit/debit records, and an integrated complaint management system. Built with React and TypeScript, featuring hierarchical user tree visualization with collapsible nodes and modal views. Backend powered by Node.js, Express, and MongoDB with Mongoose ODM for handling transactions, user management, and campaign operations. Integrated Cloudinary for image storage with automated cleanup scheduling.
Tech: Python, NumPy, Matplotlib, Tensorflow. Implemented a basic neural network from the ground up using only Python and NumPy—no external ML libraries. Designed input, hidden, and output layers; implemented forward propagation, activation functions (Sigmoid/ReLU), and backpropagation manually. Trained on sample data to demonstrate model learning.
Designed and built a backend system simulating YouTube features using Node.js, Express, and MongoDB. Implemented video upload, metadata storage, user registration & login with JWT auth. Structured REST APIs to handle likes, comments, views, and subscriptions. Deployed with Postman testing and MongoDB Atlas.
Network Marketing Platform is a full-stack web application for managing a multi-level marketing (MLM) system. It allows users to register, refer others through unique affiliate links, track team members, view transaction and purchase histories, and monitor course progress.
AI Madness is a premium web application that lets users submit a single prompt and instantly compare responses from multiple AI models (ChatGPT, Claude, Gemini, Perplexity, DeepSeek) on one dashboard. Built with Node.js, Express, MongoDB, and Cohere API, featuring glassmorphism design, neon effects, and toggle-enabled AI panels for seamless multi-model testing.
