About

The engineer behind the systems.

Omar AI is one engineer, by design — not a sales team in front of an account you'll never meet. I build AI systems that run real operations: not demos, not pilots that quietly get switched off in month three.

Muhammad Umar Farooq — AI Automation Engineer & Full-Stack Developer

Based in: Islamabad, Pakistan
Working hours: Overlapping US business hours
Education: BS Computer Engineering
Upwork: Top Rated — 22 clients, 31 contracts, 18 reviews
Focus: Legal-led — also accounting, marketing, logistics
Engagements: Fixed-price per build, or hourly ongoing

Upwork LinkedIn

About

Hi, I'm Umar — the engineer behind Omar AI.

I'm Umar. I'm a Computer Engineering graduate and a Top Rated freelancer on Upwork, and I build for any business running into the same wall: too much repetitive, high-stakes work for the team carrying it. That's included legal, accounting, marketing, and logistics operations so far — the industry changes, the problem rarely does.

What I do is narrow on purpose. I take one process that is eating a team's day — inbox triage, intake, document assembly, pipeline upkeep — and I build the system that runs it. Then I stay on while it runs, because the interesting failures never happen in week one.

I care about a small number of things, in this order: that it's accurate on your real data, that it's secure enough for work covered by privilege, and that it's still working months after I've stopped being the person watching it.

Trajectory

Now
AI automation across industries, proven on legal work
Top Rated on Upwork across 22 clients and 31 contracts, building email, demand-letter, CRM, and operations-platform automation for businesses across legal, accounting, marketing, and logistics operations.
See the work
2026
Reproducible ML pipelines
A full DVC + MLflow pipeline with a registered model and a serving API — closing the gap between an experiment that scored well and an endpoint someone can actually call.
YouTube Sentiment Insights
2025
Multi-tenant retrieval
Built and open-sourced a FastAPI RAG service with per-tenant vector isolation enforced at the storage layer — working out the confidentiality problem that every law-firm engagement runs into.
Multi-Tenant RAG API
2024
Automation and ML foundations
Early public builds: a templated bulk-email dispatcher, an LSTM climate forecaster, and a Random Forest pricing model. The email bot is the direct ancestor of the inbox systems I ship today.
See the lab

How I work

Six things I won't compromise on.

These aren't values on a wall — each one is a decision that shows up in the code, and each one costs something.

Real data or it doesn't count

Every build is tested against your actual, messy records before launch. A system that only works on the happy path is a demo, and demos don't survive contact with a Monday morning.

A human where it matters

Automation earns trust by knowing where to stop. Anything that reaches a client gets reviewed by a person unless you've explicitly decided otherwise.

Boring, durable tooling

I build on production-grade tools chosen because they'll still be here in two years — not because they trended last month. Novelty is a cost your operation pays later.

Measured, or it's an opinion

If a build claims to save time, the before and after get measured. Where a number hasn't been measured yet, I say so rather than estimating one.

One accountable engineer

You work with me, not an account manager in front of a team you'll never meet. Fewer clients, each one properly served.

Built to be handed over

The workflows, prompts, and documentation are yours. Nothing is architected so that you need me in order to keep it running.

How I operate

Remote, documented, and awake when you are.

An engagement that runs across nine time zones only works if the way it runs is explicit. This is the whole of it.

Scoped before anything is built

Every engagement opens with a walkthrough of the process as it runs today — the triggers, the tools, the exceptions nobody documents. You get a written scope with what is being built, what is explicitly out, and what it depends on from your side. No build starts from a verbal brief.

Built in your systems, on your accounts

Work lands in your Microsoft, Google, CRM and automation accounts under credentials you own and can revoke. Nothing is hosted on a personal account that leaves with me, and there is no step in any workflow that only I can run.

A person approves what matters

Anything client-facing, billable or legally consequential stops for a human. New workflows run in review-only mode first, so the team sees what the system would have sent before it is allowed to send anything.

Documented as it is built, not after

Each workflow ships with a written runbook: what it does, what it touches, how it fails, and how to switch it off. The documentation is a deliverable, not a favour — it is what makes the system survivable without me.

Watched after launch

The months after go-live are where most automation quietly rots. Runs are logged, failures surface rather than passing silently, and edge cases get fixed as they appear instead of at the next quarterly review.

Handed over to your team

The people who use the system get walked through it while it is being built, not handed a login at the end. If your team cannot operate and adjust it without me, the engagement has not finished.

Time zones

I keep US hours.

I hold the back half of my day open for US business hours. Calls are booked in your time zone, not mine, and work lands overnight so it is waiting when your day starts.

Islamabad, Pakistan — UTC+5, no daylight saving

What I will not promiseA same-hour response at 4pm Pacific is not something I will promise, because it would be the middle of my night and the promise would break. What I do commit to: a reply within your working day, and a standing call at a time that suits yours.

Time difference between Islamabad and each US time zone
US time zone	Nov–Mar	Mar–Nov	9am there
Eastern (New York)	10 hours behind	9 hours behind	7:00 pm my time
Central (Chicago)	11 hours behind	10 hours behind	8:00 pm my time
Mountain (Denver)	12 hours behind	11 hours behind	9:00 pm my time
Pacific (Los Angeles)	13 hours behind	12 hours behind	10:00 pm my time

Pakistan does not observe daylight saving, so the gap narrows by an hour when US clocks spring forward. Both are listed. The last column is the standard-time figure.

Delivery record

Verify on Upwork

Thirty-one contracts, twenty-two clients.

The full contract history behind the three case studies — every engagement, grouped by what it actually was. Client names are withheld; the work is not.

AI systems & automation

11 contracts

AI Developer for Personal Injury Demand Creation2 contracts
US personal-injury law firm
Mar 2026 – presentRunning
Automation Engineer (Make.com), long-term
AI & finance automation agency
Sep 2026 – presentRunning
AI Automation Specialist for Law Firm
US law firm
Mar – Jun 2026
CRM from Stacker & Airtable with n8n automations + AI
30-user services team
Mar – Aug 2026
Make.com automation with the Claude API
Operations consultancy
May – Aug 2026
AI Agent Developer for Nonprofit
Nonprofit
Jul 2026
AI Content Systems Manager
Content studio
Feb – Apr 2026
AI Systems Specialist for Online Course
Online course business
Jan – Apr 2026
AI Chatbot Optimization & Integration
Chatbot product team
Jun – Oct 2025
Chatbot Creation and Methodology Analysis
Research client
May – Jun 2024

Full-stack & SaaS

6 contracts

Full Stack SaaS App Development
TeaganAI
Jul 2026 – presentRunning
TeaganAI — dashboard development
TeaganAI
Mar – Jul 2026
TeaganAI — critical issues before beta launch
TeaganAI
Jan – Mar 2026
Backend software building
Software product team
Oct 2025 – Jan 2026
Python engineering, ongoing hourly
Software product team
Aug 2025 – Apr 2026
Supabase movie database tagging & AI enrichment
Media metadata platform
Apr – May 2025

Machine learning, data & vision

14 contracts

Predictive models from psychological language data2 contracts
Psychometrics company
Jul 2025
ML prediction engagements4 contracts
Research client
May – Nov 2025
ML engineer
Research client
Dec 2025 – Jan 2026
AI network (Python)2 contracts
Research client
Jan – Jun 2025
OpenCV licence-plate recognition & data extraction2 contracts
Computer-vision product company
Jul 2024 – Mar 2025
Multiple regression analysis (Python)
Research client
Jan – Feb 2026
Replicate multivariate RARCH model results
Quantitative research client
Jul 2024
Machine learning and AI tutoring
Individual
Apr 2024

Read from the Upwork contract record on 2026-10-06. Four engagements are still running.

In their words

Verify on Upwork

Testimonials

Quoted exactly as written on Upwork, each against the contract it came from. Every review on the profile is open to inspection, including the one that is not a five.

“I can't recommend this developer highly enough. He built the full TeaganAI platform — four complete dashboards, all delivered at launch with nothing deferred to later phases: 1 -School SaaS…”
Founder, TeaganAITeaganAI — Dashboard DevelopmentMar – Jul 2026 — 5.0 of 5Reliable · Committed to Quality · Solution Oriented · Clear Communicator · Detail OrientedRead more
“He is absolutely wonderful! Very grateful to have worked with him. Umar is a 10/10.”
Principal, US law firmAI Automation Specialist for Law FirmMar – Jun 2026 — 5.0 of 5Committed to Quality · Detail Oriented · Accountable for Outcomes
“Umar was outstanding throughout the project. He quickly absorbed our requirements, built a clean, well-documented classification pipeline, and iterated on improvements without prompting. His code quality, attention to data hygiene,…”
Data science team, psychometrics companyPredictive Models from Psychological Language DataJul 2025 — 5.0 of 5Clear Communicator · Collaborative · Committed to Quality · Detail Oriented · Solution OrientedRead more
“Working with Umer has been an outstanding experience from start to finish. He played a critical role in getting TeaganAI ready for beta launch, tackling some of the most important…”
Founder, TeaganAINext.js Developer for TypeScript FixJan – Mar 2026 — 5.0 of 5Reliable · Committed to Quality · Solution Oriented · Clear Communicator · ProfessionalRead more
“Umer went above and beyond to complete the task and was very helpful. He was always available to hop on a call and answer any questions, and we are delighted…”
Product owner, media metadata platformSupabase Movie Database Tagging & EnrichmentApr – May 2025 — 5.0 of 5Accountable for Outcomes · Clear Communicator · Solution OrientedRead more
“Excellent Work!”
Operations lead, 30-user services teamCRM from Stacker and Airtable with n8n Automations with AIMar – Aug 2026 — 5.0 of 5
“Umar's great, would hire again.”
Founder, content studioAI Content Systems ManagerFeb – Apr 2026 — 5.0 of 5
“Very reliable freelancer. Quick delivery and high-quality results”
Research client, prediction modelsMachine Learning PredictionOct – Nov 2025 — 5.0 of 5Reliable · Committed to Quality

Course certificates

Short courses, listed as short courses.

Completion certificates, each one linked so you can open it. They are a few hours each — the delivery record above is the evidence that matters, and these sit underneath it rather than beside it.

Separate from the security page's position on compliance: no SOC 2, ISO 27001 or HIPAA certification is held, and nothing here implies one.

Contracts delivered: 31
Faster email turnaround: ≈80%
Automations running: 24/7
On Upwork: Top Rated

Tools & stack

What the systems are actually built on.

Grouped by the job each one does. Everything here appears in work that shipped — it is a record of what I build with, not a list of logos.

Models & AI

Chosen per task, and swappable — no system here is welded to one provider.

Claude
OpenAI API
AWS Bedrock
Google Gemini
Hugging Face
PyTorch

Agents & orchestration

Where the logic lives: what runs when, what it is allowed to do, and where it stops for a person.

n8n
Make.com
LangChain
LangGraph
Webhooks

Retrieval & knowledge

How an assistant answers from your documents instead of from its own imagination.

pgvector
Pinecone
FAISS
Chroma

Backend & APIs

The services the workflows call, and the ones they expose.

Python
FastAPI
Node.js
PostgreSQL
Supabase

Apps & dashboards

Where a team actually sees and operates the system.

Next.js
React
TypeScript
Tailwind CSS
Stripe

Business systems

The tools a client already runs. The work is almost always meeting these where they are.

Microsoft Graph
Clio
Airtable
GoHighLevel
Gmail
Google Drive
Notion

Voice & messaging

For the workflows that reach a person by phone rather than by inbox.

Twilio
Vapi

Cloud & delivery

Where it runs, and how it gets there.

AWS ECS
Docker
Vercel
GitHub

Losing hours to work an AI system could handle?

Tell me about the process. I'll tell you straight whether it's worth automating — no pitch, no obligation.

Book a free audit call