What I've built
Projects
LLM-based Travel Planner
AI agentsLLM
- Designed a hierarchical multi-agent system using GPT-4o for personalized travel itineraries.
- Implemented supervisor-agent orchestration coordinating calendar, data retrieval, planning, and feedback agents.
- Built generate–evaluate–refine pipeline with rule-based scoring for constraint-aware itinerary optimization.
- Integrated APIs and datasets for flights, hotels, restaurants, and attractions with LangChain tools.
PythonLangChainLangGraphGPT-4oMulti-Agent SystemsAPI integration
AI-powered Calendar Assistant
AI agentsLLM
- Designed a hierarchical multi-agent system using GPT-4o for personalized calendar management.
- Implemented dynamic agent orchestration with distributed state management and supervisor failover.
- Integrated Google Calendar API with conflict detection, undo/redo, and real-time Streamlit interface.
- Designed horizontal scaling with Nginx load balancing across multiple supervisor instances.
PythonLangChainLangGraphNginxRedisStreamlitDocker
Real-Time Chat & Video Calling App
Full-Stack
- Built a full-stack real-time communication platform utilizing Angular 19 and .NET 10 to support instant messaging alongside peer-to-peer video and voice calls.
- Implemented P2P video and voice communication using WebRTC, configuring STUN server negotiation and ICE candidate queuing via SignalR signaling hubs.
- Developed JWT-authenticated SignalR hubs to manage real-time messaging, typing indicators, read receipts, active online presence states, and cursor-based message pagination.
- Designed a database schema to log comprehensive call metadata, capturing session duration, multi-state status tracking (completed, missed, declined, cancelled), and call direction flags per user.
AngularTypeScriptASP.NET Core.NET 10SignalRWebRTCEntity Framework CoreSQLiteTailwindCSSAngular MaterialDockerJWT
AI for IT Operations Monitoring Stack
MonitoringAI/MLAIOps
- Implemented full-stack monitoring with Prometheus, Grafana, Node Exporter, and ML-based anomaly detection.
- Developed Python ML scripts using IsolationForest for proactive anomaly detection.
- Enabled live anomaly detection, alerting, and visualization dashboards for CPU, memory, and disk metrics.
PythonDockerPrometheusGrafanaIsolationForest
- Built a full-stack gaming platform with a catalog of playable browser games including an endless runner, maze navigator, card game, and action shooter.
- Integrated Clerk authentication with a custom gaming-themed UI, supporting sign-in/sign-up modals and protected routes.
- Implemented a centralized game registry with filtering, sorting and smooth GSAP scroll animations.
ReactTypeScriptViteTailwind CSSGSAPClerkKaplayEmailJS
Corporate Portfolio Redesign - Ragvora Solutions
FreelanceFrontend Engineering
- Executed a comprehensive UI/UX overhaul of the corporate portfolio website, translating high-fidelity designs into pixel-perfect, responsive components using React and Tailwind CSS.
ReactTypeScriptNext.jsEmailJS
Kiddo Cart - School Van Service
FreelanceFull-StackMobileFrontend Engineering
- Developed a unified React Native (Expo) mobile app serving both Driver and Parent roles from a single codebase, featuring live GPS location tracking, attendance event recording, absence management, and a notification history feed.
- Engineered a NestJS REST API with role-based JWT authentication, Supabase (PostgreSQL) for data persistence.
React NativeExpoTypeScriptNestJSSupabaseZustandTurborepo
Heystream - Live Streaming Platform
Full StackFrontend EngineeringBackend Engineering
- Built a full-stack live streaming platform from scratch using Next.js and NestJS in a Turborepo monorepo.
- Implemented real-time features with Ably, live video delivery via Livepeer, authentication with Clerk, subscription/payment flows with Stripe, and creator-to-viewer chat with TalkJS.
Next.jsReactTypeScriptNestJSPrismaTailwind CSSLivepeerStripeAblyClerkTalkJSCloudinaryTurborepoPostgreSQL
Tumor.ai - AI-Powered Medical Imaging Platform
Full StackAI/ML
- Built a clinical decision support web application that empowers healthcare professionals with AI-powered detection of brain tumors, breast cancer, and skin cancer from MRI and medical images.
- Developed a responsive React/TypeScript frontend with role-based access, admin verification workflows, scan history tracking, and AI-generated heatmap visualizations.
- Integrated a Flask backend serving deep learning models (UNet, ResNet50) with Firebase authentication and MongoDB for data persistence.
ReactTypeScriptViteTailwind CSSMaterial UIFlaskPythonTensorFlowPyTorchMongoDBFirebaseONNX RuntimeOpenCV
