Projects

DevLab – Scalable Web Coding Playground

I always wondered how platforms like Codedamn or Replit let you spin up a full-stack app — React, Next.js, Go, or Python — right in the browser, isolated yet lightning fast. So I built DevLab.

What began as a simple Docker experiment soon became a deep dive into the hidden infrastructure that powers modern developer platforms. DevLab is a container-based, full-environment coding playground that spins up isolated sandboxes on demand, complete with terminals, live previews, and real-time file sync.

The Architecture

Each DevLab session runs as a Kubernetes Job, launching a Pod with multiple containers: a runtime container that executes user code, a WebSocket service for live updates, and a shared volume for instant file access. A wildcard Ingress setup dynamically routes traffic (e.g., s1.playground.domain.com) to the right environment.

I started with AWS ECS and quickly hit scaling walls — slow EFS mounts, costly runs, and frustrating feedback loops. Moving to a queue + worker model helped, but I realized I was rebuilding what Kubernetes already does better: lifecycle management, networking, storage, and scaling — all out of the box.

Building DevLab was more than just wiring containers together — it was about understanding the backbone of scalable, cloud-native infrastructure. The result isn’t just a playground, but a blueprint for running isolated, ephemeral developer environments at scale.

You can explore more about this journey in my blogs: Building a Scalable Web Coding Playground Part 1 and Optimizing the Playground Infrastructure Part 2 .

Technologies Used : Docker, Kubernetes, Redis, NestJS, Vite, Postgres

Avara - Group Video Calling Platform

Avara is a real-time group video calling application built with WebRTC and Mediasoup, designed for low-latency, high-quality video communication. It leverages an SFU (Selective Forwarding Unit) architecture, ensuring efficient media routing and scalability for multi-participant calls

Why I Chose Mediasoup?

About a year ago, I built a basic WebRTC-based video calling app. It worked—but barely. The moment more than a few participants joined, the lag kicked in, quality dropped, and the whole experience felt unusable. I knew WebRTC was powerful, but I was clearly missing something.

That’s when I started researching different architectures and stumbled upon this video. It broke down how SFUs (Selective Forwarding Units) like Mediasoup could handle large-scale calls way more efficiently than a simple peer-to-peer setup. I was hooked.

My first attempt at implementing Mediasoup? Total failure. It was overwhelming—configuring workers, handling transport, and debugging ICE issues. But I didn’t give up. I spent more time digging into Mediasoup’s API, learning how to manage media streams properly, and optimizing server performance.

On my second attempt, things finally clicked. Avara was born—running smoothly, handling multiple participants with minimal latency, and proving why Mediasoup was the right choice.

Now, Avara leverages Mediasoup’s SFU architecture to provide **low-latency, high-quality video calls** with **adaptive bitrate streaming, efficient bandwidth usage, and server-side stream control**—everything my first WebRTC app lacked.

Technologies Used : NestJs, NextJs, Postgres, Redis, Mediasoup

Nova - Edtech Platform

Nova is an innovative edtech platform that I led, focused on delivering high-quality educational content through interactive quizzes and videos. I designed and implemented a robust system for managing and distributing educational resources. Leveraging technologies like Next.js, Express.js, Redis, MongoDB, and many AWS services, I built intuitive interfaces for users and instructors, significantly enhancing the overall user experience.

Key Features and Technologies

1. Custom Video Processing: Implemented efficient video processing on our own infrastructure built on top of AWS, reducing costs compared to AWS services like Mediaconvert. Utilized FFmpeg for video transcoding, ensuring high-quality streaming.

2. Adaptive Streaming: Employed HLS for adaptive streaming, delivering smooth and uninterrupted video content to users.

3. Comprehensive Analytics: Built a detailed analytics dashboard for instructors and students, displaying key metrics such as course completion rates, engagement rates, subscription counts, and revenue.

4. AI-Powered Quiz Validation: Developed an AI-based system for automatic quiz validation, enhancing the accuracy and efficiency of assessments.

Technologies Used : MongoDB, Nextjs, ExpressJs, Redis, Aws S3, Aws Ec2, Aws Cloudfront, Aws Lambda