MADE IN KOLKATA. BUILT FOR SCALE.
I build the invisible systems that move commerce.
/ TRACKING — ONE ORDER, FIVE HOPS
HOP 01 · INGRESS · DEC 2021
Fresh out of college, straight into production. I started where data enters the system — building web scrapers for price benchmarking across the Indian market, feeding the engines behind dynamic pricing and marketing strategy.
PYTHON · NODE.JS · SCRAPING AT SCALE
HOP 02 · CATALOGUE
Then I went deeper: the product store. I helped build a distributed catalogue system flexible enough to absorb constant change — new products, new attributes, new sellers — without ever stopping the machine.
NODE.JS · POSTGRESQL · MONGODB
HOP 03 · PLATFORM
On the platform team I owned shared systems and internal tooling — and a complete, ground-up rewrite of the order service: the pipeline every single order flows through. You don't get to pause commerce while you replace its heart.
NODE.JS · KAFKA · REDIS · POSTGRES
HOP 04 · INFERENCE
With the data-science team I worked on ML-based recommendation systems — squeezing storage, accelerating data access, and turning behavioural signals into suggestions people actually wanted.
PYTHON · ML · STORAGE EFFICIENCY
HOP 05 · OPERATIONS · SDE-II SINCE OCT 2023
Software is easy; trucks are hard. Today I optimize warehouse and logistics capacity — and build the atomic guardrails (Redis + Lua) that keep distributed checkout honest under peak festival load.
GO · REDIS · LUA · CAPACITY PLANNING
/ MANIFEST — SHIPPED TO PRODUCTION
CityMall work described from orbit — the interesting details live in production.
Ground-up rewrite of the pipeline every order flows through — resilient, observable, and built to be replaced without downtime.
NODE.JS · KAFKA · POSTGRES
Race-free promotional caps for distributed checkout. Redis + Lua — atomic where it matters, because money doesn't forgive race conditions.
REDIS · LUA
The distributed catalogue behind search and discovery — schema-flexible, write-heavy, and always on.
NODE.JS · POSTGRESQL · MONGODB
Warehouse & logistics capacity planning — deciding what the network can promise, and when it can deliver.
GO · POSTGRESQL
A personal AI chief-of-staff living in Slack — approval-bus architecture, human-in-the-loop by design, and $0 when idle.
TYPESCRIPT · LLM AGENTS
An LLM-powered pipeline that turns messy cloud invoices into clean, queryable cost analytics.
PYTHON · LLMS
— ARCHIVE / EARLIER BUILDS
/ BILL OF MATERIALS
/ THE HUMAN IN THE LOOP
I'm a backend engineer in Kolkata who likes systems with consequences: orders, money, trucks, time. Four-plus years at CityMall took me through five teams — from scraping market data to running operations infrastructure — and left me with a conviction that reliability is a feature users feel.
These days I'm equally obsessed with the AI layer: building agents that do real work under human approval, and tooling that makes knowledge compound instead of evaporate.
B.TECH, COMPUTER SCIENCE — BENGAL INSTITUTE OF TECHNOLOGY (2019–2023) · CGPA 9.4
/ FINAL HOP — STATUS UPDATE
Every order ends at a doorstep. This one ends at yours — let's build something that doesn't fall over.