Megat Irfan Zackry

Megat Irfan Zackry, Software Engineer, System Architect and Automation Developer

I am a Software Engineer

I build systems and automate repetitive tasks involving bots, pipelines and platforms designed around the people who actually use them.

See my work Download resume Get in touch

01About Me

I am not trying to build technology that makes people more connected.

The internet was built to close distance between people. What got built on top of it optimised for feed ranking, notification timing and infinite scroll, all tuned to time on platform rather than time well spent. A follower count is not a friendship. A contact list is not a community.

The open question now is whether artificial intelligence finishes what social platforms started which is another automated layer between people, dressed up as connection.

I realized how it can do the opposite.

AI should not write the messages that matter or manage the relationships that matter. It shouldn’t draft your apologies, automate your check-ins, or simulate empathy to a friend in need.

Let humans keep our humanity.

Ai should take the part that was never really a relationship to begin with. The moderation queue, the spam filter, the verification email, the reminder that fires on schedule whether or not a person is the one sending it.

I am trying to build technology that makes people less busy performing connection, earning more time to actually connect with those who matter.

02What I Do

03Experience

Power Platform Developer Intern

KPMG PLT, Petaling Jaya · Sep 2023 to Feb 2024

  • Built the Smart Locker Power App, which notifies IT staff the moment an item is dropped off and replaced a manual notice process.
  • Developed the Sentinel Lead Partners reminder automation, generating personalised emails from pending action data instead of sending them one by one.
  • Maintained the firm's Expense Claim System in C# and SQL Server, plus its AI Builder document extraction pipeline running at roughly 94 percent accuracy.
  • Rebuilt the staff directory synchronisation pipeline in PowerShell, centralising active staff records across SharePoint and the firm's SQL Server source of truth.
  • Power Apps
  • Power Automate
  • C#
  • SQL Server
  • PowerShell
  • AI Builder

04Selected Work

Community Automation and Moderation Platform

Builder and operator · UTM Confession, Telegram

What started as my final year project grew into a semi-anonymous community platform serving more than 12,000 subscribers, running on a modular Google Apps Script architecture with shared libraries for Telegram API integration, caching and state management.

  • Real time moderation engine that redacts personally identifiable information, filters profanity and screens URLs before anything is published.
  • Conversational AI assistant with system instructions and prompt constraints that hold it to Telegram rich message HTML and sane behaviour across multi turn conversations.
  • Identity verification tied to university email through OTP, balancing anonymity against accountability.
  • A gated space for student entrepreneurs and clubs, with an automated approval workflow for promotions.
  • An automated marketplace for customised usernames that has generated over 2,000 Telegram Stars.
  • 12,000+ Subscribers
  • 2,000+ ⭐ Telegram Stars sold
  • 24/7 Automated moderation
  • Automated email verification
t.me/utmconfession t.me/utmchat Bot template on GitHub

End to End Booking Platform

Sole engineer · Cleaning services business

A customer facing booking site built and operated solo, from the form validation down to the mail server that sends the confirmations.

  • React with TanStack Start and Tailwind, Zod validated forms, calendar based scheduling and OTP verification.
  • Deployed on Cloudflare Pages and Workers with custom email automation and routing backed by a self hosted SMTP server, all on a single production domain.
  • React
  • TanStack Start
  • Tailwind
  • Zod
  • Cloudflare Workers
  • SMTP

05Research

Quantum Inspired Agentic Test Case Prioritisation

MPhil thesis · Universiti Teknologi Malaysia · In progress

A data independent framework for prioritising regression tests, aimed at the cold start problem: what do you run first when you have no execution history to learn from?

  • Uses large language models for semantic analysis of code changes, extracting fault risk and complexity features on the fly rather than from historical data.
  • Ranks tests through a classical quantum inspired depolarising channel model, sidestepping the computational cost of metaheuristic search such as PSO.
  • Runs on an automated GitHub Actions and Google Apps Script pipeline that pulls code diffs by webhook and triggers Python evaluation scripts.
  • Supported by a systematic literature review on the scalability and effectiveness trade offs of AI driven prioritisation.
  • Python
  • LLM evaluation
  • GitHub Actions
  • Regression testing
  • Systematic review

06On GitHub

Public repositories, mostly templates and experiments. The community platform and client work live in private Apps Script projects and private repos, so the code below is the part I can share openly.

Check out my GitHub

07Tech Stack

08Background

  • 2018

    Sijil Pelajaran Malaysia (SPM)

    8A · MRSM Tun Ghazali Shafie

  • 2020

    UTM Foundation

    3.96 CGPA · Kolej Mara Seremban

  • 2023/24

    Industrial Training

    6 months · KPMG, Petaling Jaya

  • 2025

    Bachelor of Computer Science (Software Engineering)

    3.58 CGPA · Universiti Teknologi Malaysia

  • Now

    Master of Philosophy

    Software Engineering · Universiti Teknologi Malaysia