Skip to main content

Open to senior full-stack and backend roles

Vivek Kumar

Building scalable systems, cloud-native applications, and high-performance digital products.

View ProjectsDownload Resume

600K

Variants Priced

catalogue price sync I own

4.8x

Faster Price Sync

2.6 hours down to 33 minutes

99.5%

Cache Hit Rate

two-tier master-data cache

5

Tenant Platforms

Kisna · Iraah · Sadashree +2

vivek@command-center: ~/prod
github.com/vivekumar08

3,261 contributions · 5 months

peak 1640 on Sep 21, 2026lessmore

01 · projects

Featured Builds

Production systems shipped for Kisna, Iraah, PC Jewellers and Sadashree — select a build to inspect the problem, the architecture, the trade-off, and the shipped UI.

Catalogue Price Sync Cache

~/builds/price-sync-cache

product: Kisna Diamonds · Iraah — catalogue export service

role: Owner — profiling, cache design, benchmark write-up, rollout to two tenants

Problem

Recalculating prices across a 600,000-variant jewellery catalogue took about 2.6 hours, so it consumed most of a maintenance night and could not be re-run when rates moved.

Solution

Built a two-tier master-data cache — in-process on top of Redis — with concurrent misses collapsed into one query, negative results cached so missing records stop being re-queried forever, ObjectId-safe encoding, and compression above 1KB. It opens only while a run is in progress, so imports and single-variant paths always read fresh data.

Architecture

Sync run → cache warm-up → per-variant price calculation reading product, option set, price master, meta attributes and categories through the two-tier cache → batched variant writes replacing per-document updates; batch size and concurrency are environment-tunable, defaulted to match production exactly.

Key decision

Measure before touching anything. The assumption in the room was 'it is slow because the catalogue is big'. A query census showed 14,158 lookups over 1,000 variants resolving to 143 distinct documents — the job was re-reading the same master data tens of thousands of times, and no amount of extra hardware would have fixed that.

Node.jsMongoDBRedisBenchmarkingProfiling
600K-variant run
2.6h → 33m
600K-variant run
per variant
15.82 → 3.28ms
per variant
DB queries / variant
17.95 → 0.10
DB queries / variant
cache hit rate
99.5%
cache hit rate

Shipped UI

~/shots/price-sync-benchmark.png
Benchmark: before/after on production data, with the parity checks that prove the output did not change

Benchmark: before/after on production data, with the parity checks that prove the output did not change

03 · experience

Engineering Timeline

Commit history of a career — from research internships to owning production commerce systems.

  1. Jul 2023 — Present

    HEAD → main

    Software Engineer

    Bytive Technologies — Kisna Diamonds, PC Jewellers, Iraah, Sadashree

    • Own catalogue pricing performance across tenants: profiled a 600K-variant price sync down from 2.6 hours to 33 minutes with a two-tier Redis-backed master-data cache — 17.95 to 0.10 DB queries per variant, 99.5% hit rate, output verified byte-for-byte identical.
    • Ported the same pattern to a 300K-variant tenant (78–162 min → 15–35 min) and to both Meta feed-export crons, where a query census and parity harness exposed three stacked defects including concurrent cursors doing 3x the work before deduplicating it away.
    • Designed and shipped direct browser-to-R2 uploads across three repos — presigned endpoints, a bounded-concurrency transport, and byte-level progress in the file manager — removing the app server from the upload path.
    • Took Clara, a WhatsApp commerce assistant, to production: event ingestion, Salesforce push behind an environment flag, promotions API, and a feature-flagged enquiry console for support.
    • Built event-driven SQS pipelines over 10K+ SKUs with DLQs and idempotent consumers, integrated Azure Service Bus for ERP ↔ e-commerce order sync (PC Jewellers), and cut API latency 40% moving synchronous flows onto Redis-backed queues on AWS Lambda.
    • Kept the work reviewable: 50+ test files across upload, chatbot, search and pricing paths, 7 design specs written before implementation, and benchmark documents recording methodology, variance, and the next optimisation lever.

  2. 2022

    Full Stack Developer Intern

    Daulat Ram College, University of Delhi

    • Built and shipped full-stack features across React frontends and Node.js APIs, used daily by college staff.
    • Modeled institutional workflows in MongoDB and integrated them end to end.

  3. 2021

    Full Stack Developer Intern

    Cluster Innovation Centre, University of Delhi

    • Delivered web applications from design to deployment in a research-driven team.
    • Production fundamentals under review discipline: versioning, code review, iterative delivery.

05 · blog

Engineering Notes

Writing up the systems I build — published on Medium. Queues, caches, auth, and the failure modes between them.

Distributed Systems

Designing Scalable Microservices in Node.js

A practical guide to service boundaries, contracts, and observability — designing Node.js microservices for a multi-vendor commerce platform.

4 min readNov 2025
All posts on Medium

06 · contact

Open a Connection

Hiring for a senior full-stack or backend role? Building something that needs to scale? Let's talk.

$ ssh vivek@available

Let's build something that stays up.

Open to senior full-stack and backend roles. Fastest response over email — typically within 24 hours.

India · Remote-friendly

[email protected]Resume