Senior Software Engineer · Full-stack · KL, MY

I build production-ready software. End-to-end, from idea to shipped.

8+ years shipping web apps, SaaS products, and AI tooling. I work across the full stack: spec, architecture, backend, frontend, deployment. I care about how the product lands, not just that the code runs.

Get in touch View projects GitHub CV

● Kuala Lumpur, MY · GMT+8● Snappymob, prev. enterprise SaaS● 8+ years

~/build · zsh · 80×24

engineer-build ×

spec.md

architecture.md

01 / where i help most

Four things I do well and do often.

Problems I run into constantly, at work and in side projects. I have good answers for all of them.

ISSUE-001open· engineer: @meher

Requirements exist but no one has turned them into a technical plan.

I take written requirements, user stories, or rough notes and produce a real technical roadmap: data model, API design, infra decisions, milestones, and task breakdown. Something engineers can actually build from.

roadmapspecsarchitecture

ISSUE-002open· engineer: @meher

The existing codebase is too slow, brittle, or hard to extend.

I dig into the code, find what is actually causing the problem, and fix it. Query performance, deployment pipeline, flaky data model, tight coupling. I refactor until adding a new feature does not feel like defusing a bomb.

refactorperformancedeployment

ISSUE-003open· engineer: @meher

Something needs to be built from scratch and shipped to production.

I handle the whole thing. Architecture, backend, frontend, auth, database, CI/CD, deployment, monitoring. I have done this enough times to know where projects stall and how to avoid it.

full-stackbuilddeployment

ISSUE-004open· engineer: @meher

The team wants to use AI but is not sure how to do it properly.

I work with Claude, Cursor, and agent tooling every day on real production code. I know what AI is good at, where it cuts corners, and how to set up workflows that make the whole team faster without creating a mess.

ai-toolingworkflowcode-quality

02 / build process

From idea to shipped product. No steps skipped.

Eight stages. None of them get hand-waved away. AI helps; it doesn't replace any of them.

pipeline · default-branch/ build-process.ymllast run: 2m ago · all stages passing

stage 01

Idea

Understand the business goal, users, and constraints.

stage 02

Spec

Translate into clear, developer-ready scope and tasks.

stage 03

Architect

Data model, API boundaries, auth, infra plan.

stage 04

UI

Frontend with React / Next.js. Real interactions, not mockups.

stage 05

Backend

Node.js / NestJS. APIs, jobs, integrations, data.

stage 06

Test

Jest, Cypress, integration, manual smoke. Real coverage.

stage 07

Ship

Docker, AWS, CI/CD. Observability before launch.

stage 08

Iterate

Read logs, listen to users, improve what matters.

[10:24:11]INFO discovery call · 45min · captured 12 user stories

[10:24:18]OK spec.md generated · 6 milestones · 38 tickets

[10:24:22]INFO architecture: monolith → modular, Postgres, S3, queue worker

[10:24:31]OK ci/cd: github actions → ECS · zero-downtime deploys configured

[10:24:39]WARN ai-suggested cache layer rejected · simpler memoization sufficient

[10:24:44]OK production: monitored · 99.8% uptime · ready for users

03 / signature approach

AI-assisted, human-owned engineering.

I use AI agents and CLI tools daily. They handle volume. I handle decisions. The output still has to meet the same bar as code I wrote by hand.

Product Goal

the outcome we are building toward

input

Architect Agent

system design, data model, API boundaries

auto

Coder Agent

implements modules under defined contracts

auto

Reviewer Agent

static analysis, style, anti-patterns

auto

Tester Agent

writes + runs tests, finds edge cases

auto

Human Judgment

architecture, security, trade-offs, what NOT to build

human

Production Release

shipped, monitored, owned end-to-end

Fast output, real standards.

AI handles the repetitive parts: boilerplate, scaffolds, test stubs, first drafts. I handle what actually requires thought: system design, security, trade-offs, and whether something should be built at all.

Code from an agent goes through the same review as code from anyone else. If I would not merge it from a junior engineer, I do not merge it from a model.

01.AI for repetitive work. Architecture and trade-offs stay human.

02.Every output gets reviewed. Same bar as any other code review.

03.Tests are the contract. Generated code earns trust the same way hand-written code does.

04.Boring is fine. Postgres, queues, monoliths. Chosen because they work, not because they are interesting.

05 / build artifacts

Public work. Code you can read.

A few open repos. Most client work is private. These are the projects with public code.

aiagentflow/cliopen source

CLI tool that orchestrates multi-agent AI workflows for software development.

aboutCommand-line tool for running structured multi-agent pipelines. Handles task routing, agent handoffs, and output validation so AI-assisted dev stays predictable.

Node.jsTypeScriptCommander.jsVitest

github aiagentflow.dev

raj-khan/e2specopen source

Turns project estimates into structured development specs and LLM-ready prompts.

aboutRuby CLI that takes rough project estimates and outputs structured specs, implementation plans, and prompts ready to drop into any LLM. Built on the OpenAI API.

RubyOpenAI APICLI

e2spec.com

raj-khan/seorankslabopen source

Turns Search Console data into prioritized SEO tasks and AI-ready prompts.

aboutConnects to Google Search Console and scores your pages deterministically in code. Outputs weekly reports and prompts you can hand to an LLM. GSC data never leaves your machine.

Next.jsTypeScriptSupabaseTailwind CSS

github seorankslab.com

raj-khan/nextjs-supabase-saas-boilerplatepublic

Reusable architecture and engineering conventions for Next.js + Supabase SaaS.

aboutRepo structure, auth, database, and deploy patterns I reach for when starting a new SaaS. Opinionated but lightweight.

TypeScriptNext.jsSupabaseSaaS

github

raj-khan/pre-school-keyboardpublic

A browser typing game built for one kid. Large keys, voice feedback, emoji reactions.

aboutBuilt for my own kid. Each key press shows the character in large text, speaks it aloud, changes colors, and shows a friendly emoji. Taught me a lot about input latency and audio cues.

TypeScriptWeb AudioReact

github

06 / mission log