$▋

Sourov Das

Full-Stack Developer & AI Engineer

I build full-stack web apps and AI agents. The kind that take real actions, not just chat.

Selected Work

Projects

FROM — Editorial Portfolio

Everything lives in a single fixed viewport where a scrubbed GSAP timeline coordinates animations across sections—Loader, Hero, Video, Editorial Break, Gallery, and an archive overlay. Animations are smooth, responsive, and react instantly to scroll inputs.

Next.jsReact 19Tailwind CSS v4GSAPScrollTrigger

Codebase Live Demo

Deep Dive

Problem

Static portfolios feel flat and fail to showcase visual work in a cinematic, engaging way.

Solution

A single pinned viewport where scroll progress drives a choreographed timeline of sliding layouts, text morphs, video cuts, and a dynamic interactive gallery.

Supabase Free Tier

SmartWealth — AI Finance Dashboard

SmartWealth utilizes Next.js 16 Partial Prerendering (PPR) to stream dynamic database content inside Suspense boundaries. Authentication handles Google/GitHub OAuth and standard credentials via Supabase Auth. The AI companion is connected to FastAPI backend via Server-Sent Events (SSE) and runs tool calls (like transaction creation/deletion) on your behalf.

Next.js 16React 19SupabaseFastAPIPythonTailwind CSSRecharts

Codebase Live Demo

Deep Dive

Problem

Personal finance tools often lack real-time synchronization across views and do not provide intelligent, interactive insights into budget data.

Solution

A Next.js dashboard leveraging Supabase Realtime for instant synchronization. Features interactive charts (Recharts) and an SSE-powered AI chat assistant that can perform transactions and analyze budgets.

E-Commerce Platform

A storefront with product browsing, filtering, and checkout, plus an admin side to manage it.

ReactNext.jsJavaScriptPrismaTailwind CSS

Codebase Live Demo

Deep Dive

Problem

Shoppers give up when browsing is slow and filtering is clumsy.

Solution

Built a Next.js storefront with fast product filtering and a clean checkout flow, backed by an admin area for managing the catalog and orders.

one_hub — AI Control Desk

one_hub orchestrates multiple isolated specialist agents under a central router. Each specialist agent is scoped to its own service and credentials. Write operations are proposed as pending and are only executed after user confirmation. Memory trimming keeps context costs low.

Next.js 16React 19FastAPIPythonOpenAI Agents SDKTailwind CSS v4

Codebase

Deep Dive

Problem

Switching between email, calendar, and GitHub all day causes constant context switching and noise.

Solution

A single natural-language hub using FastAPI and OpenAI Agents. All write actions (sending emails, booking calendar events) are proposed first and only run upon clicking approve.

Supabase Free Tier

EduStream LMS

A full learning workflow for students and instructors: sign-up and roles, course creation, enrollment, and per-student progress tracking. The interfaces are responsive and every form is validated with React Hook Form and Zod, so bad input is caught before it hits the database.

Next.jsTypeScriptSupabaseTailwind CSSShadcn UIZod

Codebase Live Demo

Deep Dive

Problem

Small course creators need enrollment, roles, and progress tracking without stitching five tools together.

Solution

Built authentication, role-based access, course management, enrollment, and progress tracking, with forms validated end-to-end using React Hook Form and Zod.

RAG Chatbot Platform

Create separate document collections, drop in PDFs, text, and Markdown, and query them. Each answer highlights the exact files and sections it drew from, so you can check the source instead of trusting a guess. Vector search is swappable between Qdrant and Pinecone; models between Gemini, Groq, and local Ollama.

Next.jsTypeScriptLangChainQdrantOllamaGroq

Codebase

Deep Dive

Problem

LLMs guess when you ask about your own documents, and you cannot tell what an answer was based on.

Solution

A retrieval pipeline that ingests PDF, TXT, and Markdown files, embeds them into Qdrant, and answers from the retrieved passages. Runs on local (Ollama) or cloud (Groq, Google GenAI) models and cites its sources.

How I Build

A structured approach to designing and delivering scalable digital products.

Product-First Thinking

I design solutions around real user needs and business goals.

Scalable System Design

I build architectures that grow with users and data.

Performance Optimization

Speed, efficiency, and responsiveness are core priorities.

Clean & Maintainable Code

Structured, readable, and future-proof implementation.

Reliable Deployment

Stable infrastructure with continuous integration and monitoring.

historyLearning Journey

My path to becoming a developer

JavaScript Fundamentals

2024

Started with core JavaScript concepts, DOM manipulation, and ES6+ features.

React Development

2024

Learned React hooks, state management, component architecture, and modern patterns.

Modern Web Development

2025

Built full-stack apps with Next.js and Supabase, then started moving into AI — RAG systems and agent workflows.

Building Real-World Projects

Current

Building AI agents and RAG systems, alongside a BCA at Swami Vivekananda University.

What's Next?

Interested in collaborating, discussing a project, or just connecting? Feel free to reach out.

Get in Touch

Let's Work Together

Have a project in mind or need help building a scalable web application? I'd love to hear about it.

Start a Conversation

Or connect with me directly

GitHub LinkedIn X (Twitter)