Muaz Husaini

AI engineering · agents and LLM systems · Kuala Lumpur

I build AI agents that ship to production.

IT Associate in Data & AI DevOps at Maxis, where I design and run multi-agent systems on Google Cloud.

See the workCV (PDF)

Who

Muaz Husaini, halftone portrait
Muaz Husaini
  • NowIT Associate, Data & AI DevOpsMaxis Graduate Program
  • In productionMulti-agent analyticsVertex AI, BigQuery, Cloud Run
  • Kaggle33rd of 356Poker fraud detection, top 10%
  • UKMFirst Class, 3.88Dean's List, 8 of 8 semesters

Proof

Selected work

  1. Engraved collage: a card-catalogue cabinet whose cards turn into charts, a magnifying glass and a checked shield.

    01Maxis · Data & AI DevOps · In production

    Data agents that check their own numbers

    Plain-language analytics on BigQuery and Power BI, with a guardrail that validates every figure before anyone sees it.

    • Python
    • Google ADK
    • Vertex AI Agent Engine
    • BigQuery
    Read the case study →
  2. Engraved collage: a typewriter, a stack of sealed envelopes, a pinned map, a rubber stamp and a pocket watch.

    02Independent build · Built, pre-launch

    A job-application copilot that cannot lie about you

    Every sentence it writes must cite a fact from my real CV, and a person always presses Submit.

    • Python
    • Claude Agent SDK
    • Supabase
    • Airflow
    Read the case study →
  3. Engraved collage: playing cards and poker chips under a magnifying lens, dissolving into a scatter of data points, with a ribbon medal.

    03Kaggle · Data science · Top 10%

    Competitive data science, run like research

    33rd of 356 teams in a poker fraud-detection competition, where every submission had to earn its place on an honest holdout first.

    • Python
    • DuckDB
    • LightGBM
    • PyTorch
    Read the case study →
  4. Engraved collage: a vintage ribbon microphone, a film strip, a clapperboard and a printed sound waveform.

    04Independent build · In development

    An AI video studio with a human in the loop

    Research, script and video for a Malay-language AI explainer channel, orchestrated as one pipeline, with approval gates in front of every paid step.

    • Python
    • Airflow
    • Claude
    • Supabase
    Read the case study →
  5. Engraved collage: a small wallet with receipts and coins rising out of a laptop's camera notch, beside a bar chart.

    05Independent build · macOS · My daily driver

    AI Wallet, a native notch app for the Mac

    The MacBook notch turned into a live surface for music, activities and AI usage, built in Swift and used every day.

    • Swift 6
    • SwiftUI
    • AppKit
    • XcodeGen
    Read the case study →
  6. Engraved collage: an inspection camera on an arm examining a rusty pipe joint, over an engineering blueprint.

    06PJBUMI Heavy Engineering · MLOps apprentice

    Computer vision for oil and gas inspection

    Defect detection and asset-health models for an inspection service used with oil and gas clients.

    • Python
    • TensorFlow
    • OpenCV
    • Flask
    Read the case study →
  7. Engraved collage: a café counter with a QR table card, a receipt printer pushing out an order ticket, a coffee cup and a phone.

    07Independent build · SaaS · Closed trial

    QueLess, order-ahead for Malaysian merchants

    Customers scan a QR, order ahead and pay the merchant directly. No queue at the counter, no payment-gateway fees.

    • Next.js
    • Prisma
    • Postgres
    • Railway
    Read the case study →

Also built

Method

How I build with AI agents

I run AI coding agents the way I would run a small team: a clear spec, the right model for each job, and nothing counts as done until it has actually run.

Engraved collage: an old telephone switchboard whose patch cables lead to a typewriter, a camera, a magnifying glass and a ticked checklist.
  1. 01

    Spec before code

    Every task starts as checks that must pass: a command and the output it should give. Agents build against those checks, not against a vague goal.

  2. 02

    The right model for each job

    Fast models search and take inventory, a strong model builds, and a fresh reviewer with none of the builder's assumptions tries to prove the work is not finished.

  3. 03

    Done means it ran

    Tests, builds, screenshots and real requests. "It should work" never counts, and every agent's claim is re-checked before anything merges.

  4. 04

    Guardrails in code, not prompts

    Rules that matter, such as spending caps on paid AI tools, are enforced by hooks that block the action outright. A prompt can be ignored; a hook cannot.

I also build the tools my agents use: MCP servers that connect them to image and video tools, and a self-hosted gateway for reaching other model providers.

Experience

  1. Jan 2026 to now

    Maxis · IT Associate, Data & AI DevOps

    Graduate Program. Building and running the production data agents in the reel above.

  2. 2025

    Maxis · Consumer Business Associate

    Go-to-market for Maxis and Hotlink campaigns, then pricing and subsidy analysis for Mobile Product. Automated rebate assignment for new plans.

  3. 2024

    Deloitte Consulting SEA · AI & Data intern

    Generative AI document automation on Microsoft Fabric, and business analysis for a Tableau dashboard migration.

  4. 2023 to 2024

    PJBUMI Heavy Engineering · MLOps apprentice

    Computer-vision inspection models for oil and gas clients.

  5. Dec 2024

    UKM · BSc Computer Science

    First Class Honours, CGPA 3.88. Dean's List in all eight semesters.

Contact

Need agents that hold up in production?

Open to AI engineering roles, and select freelance work through TERAS TECH.

Email MuazDownload CV

Naro and Exa, the characters from my video pipeline.