MO

Muhammad Omais

Rough problem in.
Shipped product out.

I specify it. Then I design it, build it, or lead the engineers and designers who do, whatever the project needs. AI works alongside us every day, and the tools are the easy part. What you see here is what I do: take a rough problem, give it a shape, and keep going until it ships.

Bring me a problemSee the work

What I'm made of

Code & design
live in my DNA.
AI is the enzyme.

Two strands, bonded: engineering and design. Experience across web, mobile, desktop, services and data made me an architect. Overseeing multiple projects made me a tech lead.

My experience is made of five bases.

Web, mobile, desktop, services and database. Since 1999, all five, in whatever language the job needed. Every rung of the helix is one of those five.

The age of AI

Knowing the tools is not enough. Everyone has the same models now.

What makes or breaks a project is the person guiding them: the experience to know what good looks like, the understanding of what the business needs, the judgement to pick the right tool for the job, the discipline to debug what comes back, and the ability to guide the AI, the team and the stakeholders through it.

01 · experienceI have seen this bug before.

Twenty-seven years of shipping is worth more than a faster model. I know what good looks like before the first line is written.

02 · understandingThe model knows the syntax.

I know what your business actually needs it to do, who will use it, and what happens when it is wrong.

03 · right tool for the jobNot every job is an AI job.

Sometimes it is a model, sometimes a query, sometimes a spreadsheet and a phone call. Choosing right is the first decision, and the cheapest one to get wrong.

04 · debuggingDebugging is a science, and it is getting lost.

A confident wrong answer looks exactly like a right one. I reproduce, isolate, test, and check the numbers against the source.

05 · guidanceGuiding the AI, the team and the stakeholders.

What to let the agent run, what the team must check, and what to promise the people paying for it. That is what makes or breaks a project.

Same tool. Different hands.

The same request, the same model, started at the same moment. One side follows the instruction. The other asks a few questions first. An illustration, not a benchmark.

tools only
what was asked.questions0output1 sheet
with experience
the optimal result.questions3outputchart + sheet
Tools give you exactly what was typed. Experience asks what, why and for whom first, then builds the optimal result.

Tech lead

I make the technical calls, and clear the road.

I'm the tech lead on the insurance mobile application for TruStage at Cognizant. I own the technical direction: the stack, the architecture and the standards. My job is to see the hurdles before the team hits them, and remove them when they do.

  • decide
    The stack and the architecture.

    I pick what fits the problem, not what is fashionable, and write down why, so the team builds on decisions it understands.

  • unblock
    Hurdles, cleared early.

    Integration, performance, a dependency that will not behave: I look ahead for what will stop the team, and clear it first.

  • guide
    Review, pair, raise the bar.

    The team gets better, not just the code. AI tools included: I show people how to use them well and where not to trust them.

How a project moves

projecttech leadbaseline architectureengineershurdlesfeatures MO

a hurdle comes to meI remove it and show the engineer how to handle it next time

AI, every day

Not a feature. A habit.

I use AI to write code, analyse data, run agents, research and generate ideas. Pick a session to watch an example. These are illustrations of how I work, not claims about a client. For real work, see three AI projects.

  • Claude Code
  • OpenAI
  • Google AI Studio
omais@workbench: coding

Grounded in your data

Answers come from your own records, not from guesses.

Checked before it acts

Ordinary code and people verify the work before anything irreversible happens.

Priced up front

You know what it costs to run before you commit to it.

AI projects

Where I have put AI to work.

Three projects from my company work. I cannot show the screens, so each one comes with a diagram of how it works.

5 form screens → one chat I want to take a loan What is the reason? Job lossOther reason What does this step mean? ✓ Explained. Request submitted
illustration: the same request, as a conversation
  • Tech lead and solution architect
  • Proof of concept, presented to the client

Conversational Task Assistant

Members complete a service request by chatting, not by working through a stack of forms.

The problem
To submit a service request, such as a loan request, members went through four to five separate form screens and answered many questions.
My solution
A conversational AI assistant that lets members complete the request by chatting.
How it works
The member says what they want, for example “I want to take a loan”. The assistant asks why and presents options, such as hardship reasons like job loss. The member picks an option, and the assistant gathers everything the request needs and submits it. At any point the member can ask a question, and the assistant explains that step or option in plain language.
Status
Proof of concept presented to the client. Members understand each step while they complete it, and submitting is much simpler than a multi-screen form.
What’s next
When I presented this, Jev, TypeSafe AI’s model for structured decisions, had not been released. Jev chooses from a set of options defined in advance and says how confident it is in each one. That could make matching what the member says to the right option easier and more reliable. An idea I still need to test.
page summary It asks a question The user answers ✓ interaction complete
illustration: one summary, not every element one by one
  • Tech lead and solution architect
  • Proof of concept, approved for future development

Accessibility Experience for Visually Impaired Users

A page that explains itself, so a visually impaired user understands it at once.

The problem
Standard screen-reader support, such as TalkBack, makes visually impaired users swipe through every element one by one. That does not give them a clear picture of what is on the page. Many apps treat basic screen-reader support as accessibility done, without solving the real problem.
My approach
A dedicated experience designed for visually impaired users, not a checkbox feature.
How it works
When the user opens a page, the system generates a text summary of everything on it, so they immediately understand the page. If the page needs input, the system asks the user questions, and their answers complete the interaction.
Status
Proof of concept presented and approved for future development after the main app launch. The team is deciding whether to build it into the same app or as a separate app.
review sites website email surveys AI one property, by area sentimentsatisfactionexperienceroompropertystafffront deskamenities on targetneeds attention
illustration: scores shown are not real data
  • Lead developer
  • Built

Guest Review Intelligence System

Every guest review, from every source, turned into a clear score for each property.

The problem
Guest reviews were scattered across third-party review websites, reviews on the company website and email surveys, with no clear view of how each property was performing.
My solution
An AI-powered system that gathers all the reviews and turns them into clear property scores for management.
How it works
The system collects reviews from all sources. AI analyses each review by property and sorts it into categories: positive or negative sentiment, guest satisfaction, overall experience, room cleanliness, property cleanliness, staff, front desk and amenities. A scoring layer then rates each property in each area.
Impact
Management can see which properties and areas are on target and which need attention, so they can make decisions and work directly with the properties that need to improve.

Bring me a problem

What is going wrong?

A scripted preview of how I think, not a chatbot. A demo agent will show you how it works, then it is your turn.

company projects

    side projects

      About this site

      Directed, not generated.

      I wrote the brief, designed the look and made every call. Claude Code wrote the code. One file, no framework, built over three days.

      • 10versions
      • 3days
      • 170 KBone file, measured in your browser
      • 1library: Three.js 0.160
      • 0frameworks, build steps or trackers
      • 3Drenderer: WebGL, Three.js

      tried and dropped

      changelog · v1 to v10

      10 Oct 2026

      v1from the brief

      design ›a hero that builds itself, and a diagram of what I bring to a team

      build ›one hand-written HTML file, with a terminal showreel of AI sessions

      v2the problem ideasee it on the page

      design ›bring me a problem: pick one and watch it get solved

      build ›a scripted stage that answers six kinds of problem

      v3the age-of-AI looksee it on the page

      design ›dark aurora, glass panels, headings that stream in like tokens

      build ›a particle field that forms a problem, a system, a product and a check, and a demo agent that tries the tool first

      v4the pillarssee it on the page

      design ›five pillars: knowing the tools is not enough

      build ›two sessions racing side by side: same tool, different hands

      v5DNAsee it on the page

      design ›experience as a double helix: an engineering strand bonded to a design strand

      build ›five genes that drive the helix, and bases you can light up

      v6the ribbon helix

      design ›solid ribbons with lettered base pairs, and a fifth base: services

      build ›the hero and the helix split into sections of their own

      v7real 3Dsee it on the page

      build ›Three.js: a particle cloud with depth, an instanced-mesh helix, cards that tilt under the pointer

      11 Oct 2026

      v8one screen for the genes

      design ›the five genes as tabs on one screen, and four facts on the first screen

      build ›one helix instead of two, and a fix for reduced motion

      v9the tech leadsee it on the page

      design ›rewritten around the tech lead role: the stack, the architecture, the hurdles

      build ›a lead diagram where each hurdle clears before the work goes live

      15 Oct 2026

      v10this pagesee it on the page

      design ›three AI projects from my company work, each with a diagram of how it works

      build ›SVG diagrams that animate in CSS and hold still under reduced motion

      Contact

      Got a problem worth solving?

      Tell me what's not working. I'll reply with an honest view of how I'd approach it.

      Email Omais
      • Higher DiplomaComputer Sciences
      • CertificationsCCNA, MCSE
      • BSc, ComputingStaffordshire University, UK
      • MBA, Management Information SystemsNewport, Pakistan
      • BA, Graphic ArtsBucks New University, UK