Ehmad
Zubair

Before Ehmad Zubair sold a single AI automation to anyone else, he ran the experiment on Cogent Labs itself. Headcount: 87. Post-restructure: 58. Output: unchanged. The 29 roles AI-assisted workflows absorbed weren't a business failure - they were a proof point. That's the foundation of what he now sells: not software, but the lived conviction that AI can carry operational load without the lights going out.

Today he runs Cogent Labs from Lahore, Pakistan, with 86 people and a client list that includes UNICEF, the NHS, and a sprawling portfolio of SMEs across finance, healthcare, logistics, and retail. The company he co-founded in 2021 with Ahmed Buksh and Sher Ali has never taken external funding, never chased a flashy pivot, and has maintained a 100% on-time, on-budget delivery record since 2023. In an industry defined by missed deadlines and scope creep, that number is either very boring or very remarkable. Ehmad would say both - and mean it as a compliment.

His career started in 2014, with freelancing gigs on oDesk - the platform that became Upwork - and a degree in Information and Communication Systems Engineering from NUST (National University of Sciences and Technology) in Islamabad. That early freelancing habit of treating every client as a real business problem, not a ticket number, bled into everything that followed: iOS engineering at edX, the MIT/Harvard-backed online learning platform; software lead at Hypernym FZ LLC; senior engineer at Arbisoft, where he led teams delivering flight search and ticketing platforms at scale.

By the time he left Arbisoft in early 2021, Ehmad had spent nearly a decade watching software get built right and wrong. He had seen projects over-engineered, under-tested, and under-delivered. He had also seen what happens when curious, adaptable engineers get clear requirements and genuine ownership. Cogent Labs was a bet that the second version was repeatable.

The company's positioning is deliberately unglamorous. Where competitors pitched "transformative AI experiences," Cogent Labs published a price list: a $3,000 AI Audit that maps your operations and hands back a prioritized roadmap; an AI Build engagement running from $15K to $75K for production-ready automation. No pilots that live in slide decks. No proofs-of-concept that never scale. The tagline - "boring AI that actually works" - is the entire pitch, compressed.

Clients show up with problems like: too many hours spent on data entry, status updates that require three people to produce, report generation that eats half a workday. Cogent Labs builds workflows that fix those specific things, in weeks, on budget. The case study isn't a glowing testimonial - it's a number: hours reclaimed per week, cost per output reduced, headcount redeployed to higher-value work.

There's a version of Ehmad Zubair that never founded a company. Just a guy from Lahore who graduated from NUST, got lucky enough to land good engineering roles, and kept the knowledge to himself. That version doesn't exist. Instead, he built a second career in public.

The Desi Career Podcast - "desi" being the South Asian colloquial for "from our homeland" - started from a single sentence of personal frustration: nobody guided him through a computer science degree in Pakistan. Not a mentor, not a career counselor, not a senior engineer who took a lunch meeting. He bootstrapped a path through freelance gigs and nanodegrees and self-directed reading. Then he turned that path into a show.

Episodes cover the questions students don't know to ask: how to build an open-source contribution record before graduation, what it actually takes to get hired at a Pakistani tech firm, whether a generalist or specialist track makes more sense in Lahore versus remote-first roles. The guests are engineers at Arbisoft, VentureDive, and Noon. Alumni of NUST, FAST, and UET. People talking in the register of a direct message, not a keynote.

The YouTube channel runs parallel to the podcast but goes deeper on craft. Python bootcamp sessions. Career roadmaps. Industry analysis - including a pointed conversation about whether Pakistan's decade-long IT boom is structurally equipped to survive the AI tsunami or whether it was always a labor arbitrage story waiting for automation to arrive. Ehmad asks that question out loud, which is the kind of thing a comfortable CEO doesn't usually do.

In a YouTube conversation titled "Pakistan's IT Industry vs. The AI Tsunami: Golden Decade Over," Ehmad asks a question that comfortable tech founders usually avoid: what happens to a country's software industry when the labor arbitrage that powered its growth gets automated?

Pakistan's IT boom was built on skilled, cost-competitive engineers. That's a structural advantage - until the tasks those engineers performed become model-callable. Ehmad doesn't pretend there's an easy answer. His contribution is doing the thing that matters when the answer isn't clear: building companies that deploy AI practically, creating content that helps students develop the higher-order skills that survive automation, and asking the question publicly so someone else can argue with him.

On the enterprise side, he's direct about the friction. Data privacy concerns are the main obstacle when selling generative AI to larger organizations. Enterprises move slowly on AI adoption not because they lack ambition but because they can't afford the regulatory and reputational cost of getting it wrong. Cogent Labs' response isn't to fight that objection - it's to build around it, starting with smaller, high-ROI automations that demonstrate the methodology before touching sensitive data pipelines.

He's also building AI fact-checking agents to reduce hallucinations in production deployments - not because it's fashionable, but because a client whose AI tool confidently produces wrong answers will cancel a contract. The unglamorous work is the product.