About

A bit about me — my background, work, projects and the tools I use daily.

Umang Kumar

I'm a software development engineer at Meesho, one of India's largest e-commerce platforms, where I build the backend systems behind returns and logistics. Right now that means an event-driven reallocation pipeline moving over a million orders a day, and a dynamic rule engine that lets 25+ teams ship business-logic changes without a deploy.

I studied computer science at NIT Delhi, and competitive programming shaped how I think about code — I'm a Guardian on LeetCode, an Expert on Codeforces, and once cracked the global top 400 in Google Kickstart.

I care about systems that stay simple under load — config-driven, idempotent, observable — and code the next person can read. The full story: my work history and the tools I use every day, is below.

Work

M

Meesho

Software Development Engineer II

Jul 2025 – Present

Designed the Dynamic Rule Engine — a client library with runtime rule evaluation, caching, kill-switches and traffic-level experimentation — enabling zero-deploy business changes, now adopted by 25+ teams. Led 3PL integration and the end-to-end revamp of return reallocation into a config-driven, idempotent, event-based pipeline handling 1M+ orders a day, cutting duplicate actions by 90% and lifting returns pickup by 95%.

M

Meesho

Software Development Engineer I

Jul 2024 – Jun 2025

Owned pre-order guardrails across 520M+ product IDs, designing an incremental, change-only sync strategy that kept pricing-stack latency within SLA. Delivered upfront delivery dates for exchange orders through a low-latency API with graceful fallback, and re-architected the legacy returns approval workflow into a new service with modular boundaries and idempotent processing.

M

Meesho

SDE Intern

Jan 2024 – Jun 2024

Built the branded-packaging relinking feature from scratch — now serving 15% of BP requests — streamlined vendor onboarding to need zero code changes, and raised unit-test coverage by 40%.

K

K2Q Capital

Research & Development Intern

Jun 2023 – Aug 2023

Developed a CNN model predicting future stock performance at 80% accuracy, and built an efficient stock-analysis program for large datasets with customizable plotting.

Outside work I compete: winner of the Salesforce Codathon at Cognizance '21 (IIT Roorkee), finalist at the Bosch-sponsored Encode '23 (IIT Guwahati), Guardian on LeetCode with a max rating of 2252 and Expert on Codeforces at 1675. There's more on my LinkedIn.

Education

I hold a B.Tech in Computer Science and Engineering from the National Institute of Technology, Delhi (2020 – 2024).

Stack

  • JavaPrimary language for services at Meesho.
  • GoFast, small services and tooling.
  • C++Competitive programming weapon of choice.
  • PythonScripts, analysis and ML experiments.
  • JavaScriptThe web bits.