Skip to main content

<abhimvp.dev /> - v2.0

Hi, I'm Abhishek.

Full Stack Developer

Building scalable apps and exploring autonomous AI agents.

// 01 - About

Crafting digital experiences

I'm Abhishek (Abhi) - a Full Stack Developer passionate about building things that live on the internet. I specialize in React & Next.js while exploring the frontiers of AI Agents with LangChain and LangGraph.

Beyond code, I'm a storyteller at heart. I run on chai and movies, and in my free time, I write scenes and stories - crafting narratives I hope to one day bring to the big screen.

When I'm not coding or writing, you'll find me drawing inspiration from Max Verstappen's relentless drive, Virat Kohli's discipline, and the sheer scale of a Prabhas movie.

I use AI tools for rapid prototyping, but every architecture decision, design choice, and line of production code has my fingerprint.

Tech

ReactNext.jsTypeScriptPythonNode.jsGoLangChainLangGraphTailwind CSSPostgreSQLAWSDocker

GitHub Activity

View full profile →

// 02 - Services

What I build

From concept to production. Precision engineering, scalable architecture.

01

Custom Development

Bespoke web applications built from the ground up. Modern frameworks, scalable architecture, and pixel-perfect execution.

02

System Modernization

Transform legacy systems into modern, high-performance digital experiences. Better UI/UX, improved SEO, and future-proof design.

03

AI Integration

Leverage cutting-edge AI to supercharge your product. Chatbots, intelligent automation, and smart features that learn and adapt.

// 03 - Innovation

Building in Public

Personal OS

I wanted one interface to replace 5 apps - finances, movies, journals - all through natural conversation. No forms, just talk to it.

3 data nodes•Zero-form UX•Sub-200ms

Visual demo coming soon - currently polishing the UI.

LangGraph Agent
View on GitHub

LangGraph Agent Core

A StateGraph engine with conditional routing, parallel fan-out via the Send API, and interrupt() for human-in-the-loop confirmation. One message like "I watched Interstellar and spent ₹800 on dinner" triggers multiple nodes simultaneously.

Multi-Intent Routing

Gemini classifies intents via Pydantic + Literal types into structured output.

Human-in-the-Loop

interrupt() pauses before writes. Each branch gets its own interrupt ID.

Generative UI

push_ui_message() renders typed React components directly from graph nodes.

Parallel Fan-Out

Send API dispatches nodes in parallel. add_messages reducer merges results.

Stack:LangGraphLangChainGeminiReactFastAPISQLite

Mind & Media

Journal Node

Built

Extracts content, mood, topic tags, and reminder intervals. Saves instantly without confirmation - low-stakes writes don't need friction.

Movie Node

Built
  • •Mood Tags: "Emotional Climax", "Inspirational"
  • •Generative UI: MovieLogCard rendered from graph
  • •Interrupt: Pauses for confirmation before save

Finance & Dashboard

Finance Node

Built

Extracts amount, currency, category, description, and type from natural language. Pauses with interrupt() before committing to SQLite.

Dashboard

Built

Three clickable stat tiles open modal overlays with full data views. Powered by a separate FastAPI server with REST endpoints.

What's Next

Planned
  • Chroma memory layer for learning user patterns
  • Trip log sub-graph with cost summaries
  • LangSmith tracing for full observability
  • Natural language queries & soft delete / undo