Abhishek Kumar

Software engineer in Gurugram. Four-plus years of React, Next.js, and TypeScript in production, currently on large e-commerce storefronts. Outside work I build small open-source tools for AI coding agents, most recently agent-fastpath.

Open to opportunities

View the work GitHub

Zenex · Frontend · 2021—23 theoceann · Frontend · 2023 Infyni · SDE · 2024—25 Forchunex · Magrabi / Doctor M · now DevVault · 2026 PIE · side project · 2026 agent-fastpath · 2026

Career timeline (year)

Years shipping

since Oct 2021

agent-fastpath

v0.2.0

on npm · MIT

Platforms in production

e-commerce · edtech · maritime · health

MCP directories

MCP Market · Glama

Where the signal shipped

Four roles, four industries, one direction: from maintaining frontends to owning them, and from consuming APIs to building the platforms around them.

Forchunex IT Consultancy

Software Engineer

Dec 2025 — present · remote

Building large-scale e-commerce for Magrabi and Doctor M, two of Saudi Arabia's leading optical and healthcare retail brands.
Working across the Next.js storefront end to end: catalog, search, checkout with multiple payment methods, and integrations like virtual try-on.
Microservices platform behind it: GraphQL APIs, Akeneo PIM, Sanity CMS, Elasticsearch and Algolia search, MongoDB, ERP integrations.
CI/CD on Azure DevOps; AI-assisted engineering workflow (Claude Code, Cursor) as a daily practice.

Next.jsTypeScriptGraphQLMicroservicesE-commerce

MagrabiLive · sa-en

Doctor MLive · sa-en

Infyni

Software Development Engineer

Jan 2024 — Nov 2025 · Gurugram

Built and optimized the live-learning platform's web experience in React and Next.js.
Worked across both products: infyni.com and infyni Connect, the virtual-classroom platform for schools, institutes, and tutors.
Role-based dashboards for students, instructors, and business admins; interactive live-class flows.
Mentored teammates on hooks, component architecture, and modern React practice.

Next.jsReactTypeScriptEdTech

InfyniLive · infyni.com

infyni Connect virtual-classroom platform homepage

infyni ConnectLive · infyniconnect.com

theoceann.com

Frontend Developer

Apr 2023 — Nov 2023 · Noida

Upgraded the maritime platform to React 18; migrated class components to functions with the team.
Implemented real-time chat over WebSockets; notifications and logging with Firebase; JWT authentication.
Managed the Jenkins CI/CD pipeline for development and deployment.

React 18WebSocketsFirebaseJenkins

Zenex Animal Health

Frontend Developer

Oct 2021 — Mar 2023 · remote

Enhanced and maintained the employee reporting system's frontend in React.
Integrated against backend APIs and databases across the reporting workflow.

ReactREST APIs

Open source

Small tools I build and maintain outside work.

agent-fastpath

An MCP server that gives coding agents like Claude Code, Codex, and Cursor quick checks before they act: is this CI log ready to ship, is this command destructive, which of these files matter for the task. Clear-cut cases are decided by plain rules; the rest go to TypeSafe's Jev API. When it isn't sure, it says so and hands the decision back.

ToolsEvaluate (ship gate, risk, ambiguity presets), file triage, a headless browser check, and an evidence trace for every answer.
ResultsEvery answer is accept, review, or escalate, so the agent knows when to ask you.
ContextTriage reads files on the server; in the repo benchmark the agent read 408 tokens instead of 35,256.
StackTypeScript, Node.js, Model Context Protocol SDK, Playwright. MIT licensed.
Listednpm, MCP Market, and Glama. Setup guides for 12 MCP clients.

Website GitHub npm MCP Market

How a check is decided

Agent asksCI log, shell command, file list, or page claim

Rules firstobvious cases decided in code, no API call

Jevtyped answer with a probability

Confidence gatelow or near-tied answers are not trusted

accept · review · escalatereturned to the agent with evidence

Add to Claude Code

claude mcp add agent-fastpath -s user -e TYPESAFE_API_KEY=<key> -- npx -y agent-fastpath start

PIE

Side project · Tauri · whisper.cpp · macOS

Personal Intent Engine: a desktop app where you press a hotkey, speak, and get a structured prompt at your cursor. Transcription runs locally. Built to learn Rust and on-device speech.

Repository