Arya. Sarkar.

About

Building dependable systems.

I am a Computer Science & Engineering graduate from IIIT Manipur, specializing in Artificial Intelligence & Data Science, with 6 months of Software Development Internship experience at Delhivery. During my internship, I worked on backend systems, microservices, CI optimization, pre-production environments, and observability in distributed systems.

My technical skills include Java, Go, Spring Boot, Docker, Kafka, Redis, PostgreSQL, and MongoDB, with experience building backend and full-stack applications.

01 / Environment

Built an isolated QA pre-production environment for a distributed system of 25+ microservices by reusing shared AWS staging infrastructure and namespacing Kafka topics and Redis keys; reduced staging drift by 1000+ commits and enabled safer parallel QA testing.

Full context

The work separated QA traffic from a staging environment shared by more than 25 services without duplicating the entire AWS stack. Kafka topics and Redis keys were namespaced at the environment boundary, allowing teams to validate changes against a stable code line while staging continued to move independently.

02 / CI systems

Improved CI efficiency across 18+ Java microservices by moving shared dependencies into prebuilt Docker base images and adding Git-diff-based build gating, reducing Docker build time from 8 minutes to 1.5 minutes for service-level changes.

Full context

The CI path was optimized at two levels: stable shared dependencies were moved into versioned base images, while Git diff analysis determined which services actually required a new build. Together these changes reduced repeated dependency work and shortened feedback for service-scoped changes.

03 / Observability

Stabilized pre-production microservice flows using CubeAPM and standardized request-ID propagation across Java, Go, and Python/Django services, improving traceability across distributed request workflows.

Full context

Request identifiers were normalized across Java, Go, and Django service boundaries so a single workflow could be followed through logs and traces. CubeAPM then provided a consistent view of failures in pre-production, reducing the time spent correlating symptoms across services.

01 / 03

01 / Backend infrastructure

February 2026

FatURL High-Throughput URL Shortener & Analytics

Personal Project

Go
Spring Boot
Kafka
Redis
PostgreSQL
TimescaleDB
Traefik
Keycloak

A high-throughput redirection and event-driven analytics service with Redis-first caching, reliable Kafka publishing, and distributed ID generation.

What it does

Resolves shortened URLs through a cache-first Go service while asynchronously collecting click analytics without adding latency to the redirect path.

Why I built it

To explore the failure modes behind apparently simple redirect systems: hot keys, cache stampedes, reliable event delivery, retry safety, and horizontally generated identifiers.

Architecture

Redis uses a four-hour TTL and singleflight request coalescing. A transactional outbox publishes to Kafka, backed by a 30K-event memory buffer and idempotent consumers. PostgreSQL stores URL metadata and TimescaleDB handles time-series analytics.

sub-100ms p99 · 80%+ cache hits · 32K+ IDs/sec

02 / Developer tooling

November 2025

Code Review & Execution Platform

Academic Project

React.js
Node.js
Socket.io
Docker
QEMU/KVM
OpenAI API

A browser-based project review workspace combining file management, isolated execution, real-time terminals, chat, and an AI coding assistant.

What it does

Gives students and reviewers role-specific dashboards, project upload and extraction, a Monaco-based editor, xterm.js terminals, chat, and live development previews.

Why I built it

To remove the environment and tooling gap between a student submitting a project and a reviewer trying to inspect, execute, discuss, and assess it.

Architecture

Socket.IO carries collaboration and terminal streams. Execution sessions run in constrained Docker containers or cloud-init QEMU/KVM machines with copy-on-write disks, SSH access, dynamic preview proxying, and automated cleanup.

03 / Android product

October 2025

Lumen Android Event Management App

Academic Project

Kotlin
Jetpack Compose
Firebase
Hilt
CameraX
ML Kit

A real-time, role-aware event application for guest approvals, QR check-in, event media, and organizer-controlled content.

What it does

Coordinates event operations from invite-code joining and guest approval through UUID-backed QR admission, shared galleries, and embedded playlist playback.

Why I built it

To model a mobile product where organizer and guest experiences share live data but require different permissions, workflows, and visibility controls.

Architecture

The app uses MVVM, Hilt, StateFlow, and Firestore snapshot listeners. CameraX and ML Kit power QR scanning, while Firebase Storage media uploads include compression and EXIF-aware rotation correction.

01 / 03