👋 Hi there, I’m Janak
I’m an all-rounder in the tech world—passionate about AI/ML, full‑stack development, and DevOps. Currently, I'm channeling my curiosity into a meme generator app powered by MCP servers.

What I bring to the table:

AI/ML Enthusiast: Building intelligent systems, experimenting with models, and always learning.

Full‑Stack Developer: Crafting clean, responsive frontend UIs and robust backend systems.

DevOps Pro: Automating deployments, managing servers on Linux, writing bash scripts, and orchestrating infrastructure.

GitHub + Linux + Bash: Command‑line veteran sharing tools and projects with the community.

I love diving deep into code, deploying scalable systems, and pushing the boundaries of AI-driven creativity. Let’s connect!

• Built a full-stack agentic email system where AI can search, read, draft, and send emails across a live Gmail workflow using Next.js, FastAPI, and LangGraph.
• Engineered CopilotKit AG-UI so the agent doesn’t just respond, it drives the interface by navigating inbox states, opening threads, and preparing actions in real time.
• Shipped a production-style backend with secure Google OAuth, typed APIs, natural-language-to-Gmail search, caching, pagination, and attachment-aware email flows.
• Designed a human-approved autonomous agent with Gemini + LangGraph that keeps state across conversations, chooses tools dynamically, and pauses before every high-impact action.

• Built a 3-service AI content pipeline using Next.js, FastAPI + LangGraph, and FastMCP to automate meme-to-video generation and publishing.
• Implemented a 5-step workflow covering meme fetch, Gemini-based metadata generation, 1080x1920 video rendering with MoviePy, human approval via interrupt/resume, and YouTube upload with OAuth.
• The system spans 72 non-.git files, around 4.7K lines of code/config, 6 frontend pages, 3 API endpoints, and 5 MCP tools.
• Demonstrates agentic orchestration, Redis-backed checkpointing, multimodal AI integration, and full-stack automation in a production-style architecture.

• Built a full-stack lottery platform using React, Vite, Express, Prisma, and PostgreSQL.
• Implemented real-time style result updates, local caching, and date-based result history for an interactive game flow.
• Designed a 2-minute round-based result engine with time validation and duplicate-result prevention.
• Added JWT authentication, API-key-secured APIs, and protected frontend routes for secure access control.
• Handled transaction-safe record creation and wallet deduction to model balance-sensitive backend operations.
• Showcased experience in full-stack architecture, scheduled systems, business logic, and state-driven UI development.

Shows weather data such as temperature, wind, humidity, etc. A very basic React Project.

It detects counterfeit products in s block-supply chain. There are 3 main users manufacturer, retailer, and customer. Our app monitors the supply chain and if any mishaps or illegal activity is detected, it sends an alert to the responsible parties and also shows this data to the customer.

Just a custom social media app that utilizes React-Redux.Similar to Instagram but better and cleaner than it.

Link for Frontend code - Link Link for backend code - Link

Uses the LangChain framework to make AI development easier. We use ChromaDB for storing and retrieving embeddings. Basically, we create embeddings (vector data) and then based upon user query fetch related text extracted from pdfs and then send them to OpenAI for processing. After this processing, the ChatBot gives us the required answer.