Sumeet Umbalkar

Full-Stack & Systems Engineer

I design and build backend-heavy, real-time systems designed to scale under active production load. My focus centers on high-concurrency challenges: synchronizing state across live clients, isolating bottlenecks, and keeping system architectures stable as traffic spikes.

My daily stack centers on JavaScript and the MERN stack for core product features, WebSockets for instant state updates, Redis for cross-instance pub/sub, and Docker for reproducible deployments. I also leverage C++ and Go for algorithmic problem-solving and foundational computer science principles.

I take a systems-first approach: model the authoritative state, plan for failure modes, lock down the data structures, and then build the client interface on top of a resilient foundation.


Work

Tambola

epictambola.com · Freelance · Lead Engineer & System Architect

A multiplayer Tambola (housie) platform built from scratch to production for a client, supporting multiple simultaneous rooms with high concurrent player density and customizable game rules. I owned the architecture, design, and implementation end to end — from state synchronization models to server infrastructure deployment.

The core engineering challenge was handling real-time claim validation and state updates without bottlenecking Node.js's single-threaded event loop. I structured room, ticket, and number operations using maps and sets, keeping critical claim checks at O(1) time complexity, a few secondary operations at O(K), and only basic bulk routines reaching O(N). Game state broadcasts are strictly room-scoped over WebSockets to minimize network overhead.

To determine exact hardware boundaries, I load-tested the production cloud environment under worst-case concurrency spikes (500 active players per room executing simultaneous number draws and win claims). Performance metrics showed a single CPU core sustaining 4 rooms at a 90% peak CPU spike and ~1.4 GB RAM. Based on these benchmarks, I established a safe production ceiling of 2–3 fully loaded rooms (1,000–1,500 active players) per core instance.

Node.js WebSockets Socket.IO Redis Docker React O(1) Data Structures Load Testing

Labgineer

labgineer.com · Founder & builder

A platform I built and run for selling engineering final-year projects, spanning IoT and software, with R&D behind each listing. Built solo on the MERN stack — including the storefront, an admin dashboard for managing listings and orders, and the underlying hardware/software documentation for every IoT build. 14+ projects sold to date.

A few of the listed builds:

  • 1. G2V / V2G (Grid-to-Vehicle / Vehicle-to-Grid) Bidirectional EV Controller
  • 2. Trail Tracker
  • 3. Laser Communication System
  • 4. Smart Dustbin (Dry, Wet & Metal Sorting)
MongoDB Express.js React Node.js IoT

Open Source

liwebjs · docs

A real-time communication framework built on WebSockets for Node.js, created as my BCA final-year project. Provides rooms, channels, in-memory state, and auth, with a developer experience similar to Socket.IO — built to understand what a library like Socket.IO actually has to solve under the hood, not just to consume it.

PayGuard · docs

An agentic payment security and orchestration runtime for Razorpay, designed to let AI agents execute payments safely through deterministic policies, behavioral analysis, idempotency controls, and human approval workflows.

Memovable memovable.tech · not released yet

Context that LLMs love: open-source, npm-packaged, and MIT-licensed. Memovable is a Node.js library that gives an relevant context of existing memory based on a query to an LLM.


Projects

RivalFree · https://rivalfree.sumeet.app · Hackathon Project

An autonomous competitive intelligence and feature discovery platform built with the MERN stack. Uses multi-agent LLM orchestration and live web search to analyze competitors, identify feature gaps, and generate actionable product strategies and specifications.

More projects →

Resume

View or download my resume in PDF format.


Approach

I treat state and data structures as the first design decision, not an implementation detail — picking the structure that makes the hot-path operation cheap (usually O(1) or O(log n)) before writing any UI. I default to designing for horizontal scaling early: keep services stateless where possible, push shared state into Redis, and containerize with Docker so the same image runs the same way in dev and in production. And I try to understand the primitives underneath the frameworks I use — building liwebjs and liqueue from scratch is part of how I keep that understanding current, alongside using C++ and Go when I want closer control over performance and concurrency.


Skills

Languages

JavaScript, TypeScript, C++, Go

Backend & Real-Time

Node.js, Express.js, WebSockets, Socket.IO, Redis

Frontend

React

Data & Infrastructure

MongoDB, Docker

Focus areas: real-time system design, scalable backend architecture, and production deployment.


Education

MCA — Vellore Institute of Technology (Online)

2026 – 2028 · in progress

BCA — Tilak Maharashtra Vidyapeeth

2023 – 2026 · CGPA 7.9