Raheel Siddiqui built Pakistan's AI gov hub in half an hour

The America.gov moment, and a Pakistani answer

Muzamil opens with the launch that started it all. The US government had unveiled a central AI hub, a single place where any American could ask about any government service. It was launched with a polished, Apple-style stage show. Muzamil had the reaction many Pakistanis have to such news: what would happen if Pakistan had this? He assumed nobody would build it, since the government “is not so progressive.”

A couple of days later, someone tagged him in a post about Raheel Siddiqui. Raheel had built the same idea for Pakistan: pakistangov.online. Muzamil demonstrates it on air. He asks how to renew a Pakistani passport while living in Atlanta. The tool reads through the government’s old websites, pulls out the relevant details, and answers with sources. The site states clearly that it is independent, not an official government website.

Muzamil introduces Raheel as a 25-year-old with “no connection with the government,” just a technical person who thought, “if I can do it, why not.”

Half an hour with Claude

Raheel tells the story simply. He was working on his own product when a non-technical friend messaged him on WhatsApp about the American site and said it would be great to build one for Pakistan. Raheel told him it could be built with Claude in half an hour.

So he did it. He gave Claude the outline and some links from the government websites, then went back to his main work. Half an hour later he had a deployed link. He sent it to his friend, who was “blown away.” Raheel then added more to it, put it online, and shared it. It blew up.

Running it on almost nothing

Muzamil notes that one of Raheel’s posts was seen by close to a million people, and that the site faced DDoS attacks. He asks how an unfunded project survives the cost of every AI query.

Raheel says he kept it simple on purpose. He chose not to add a login, because sign-up would be a hassle for users. Instead he relied on lighter controls: a device asking more than about ten questions gets no more answers, since almost nobody asks that many in a day. Within three days the site had more than 3,000 users. One message stood out to him: a person who had moved to America struggled with a child’s passport, and the tool gave accurate, sourced answers.

On cost, Raheel noticed on day one that many people ask the same questions. So he cached repeat answers in a database instead of calling the AI model each time. A Karachi passport question may get a fresh answer, but a common Lahore one may come straight from the cache. The AI cost runs on Google Cloud’s free 300 dollars of credit. He plans to keep it running as long as he can.

Offers from strangers, and an IT minister

The response has surprised him. “Everyone has been so nice and kind,” Raheel says. Many overseas Pakistanis have offered to help. The IT minister also reached out, and Raheel is now talking with them about where the project could go. Companies have been pitching him salaries too.

Muzamil draws the lesson. A technical viewer might think Raheel just stitched a few things together. That is the point. “Creating products is no longer the difficult part anymore,” Muzamil says. What matters is “the depth of your imagination, the speed at which you can execute,” and designing it “tastefully” so people find value. He contrasts this with a man he spoke to the day before, depressed about losing his job, while Raheel was turning down offers after a half-hour build. The difference, Muzamil says, is that Raheel validated his skills “in the open public.”

From a Karachi classroom to building with AI

Raheel graduated in 2022 from Usman Institute of Technology in Karachi. He joined software engineering “by accident” and was bad at programming at first. In his second semester, an object-oriented programming course got him interested. He joined a university research lab working on an open-source microprocessor for Pakistan, which Google later helped put out.

He felt his university lacked the tech and founder clubs of bigger schools, so he built that culture himself. He joined a Google student program and hosted events. He applied to the GitHub Campus Expert program and was rejected. So he built an app that tells you who unfollowed you on GitHub, a running joke in tech circles, and posted it on Twitter. Someone from GitHub saw it and invited him in. He calls this his “back door strategy.”

He worked part-time for a Karachi travel-booking platform and for a US stock and crypto analytics company. His final-year project used BERT, so he understood the ideas behind large language models before ChatGPT arrived. After graduating, he spent two and a half years at a Singaporean company, then a year at an AI company in the Netherlands, where he worked on projects with the Dutch government. He built a product with Cursor, sold it, travelled, and then co-founded a real estate sales platform for the US market. It reached real revenue, but the founders could not work at the same pace, so he left. He is now building his own product.

Advice for computer science students

Muzamil frames the anxiety. When he graduated in 2015, computer science was a second choice. By 2020, everyone wanted it, for remote dollar jobs. Now the market has flipped. Is today’s CS student screwed?

Raheel calls it “an elephant in the room.” Even before AI, he says, the fastest-rising engineers were never the best typists of code. They were “really good problem solvers.” Students still need the basics, programming fundamentals, object-oriented programming, and data structures. His advice is to cut off AI while learning them, make it fun, and learn through projects. His own teachers ran project-based learning every semester.

The bigger risk, he says, is procrastination. Everyone thinks an idea will take half an hour, so nobody ships. His tool took under an hour, but “I was able to put it out there.” Students need problem solving, technical basics, and “aesthetics and taste of building a good product.”

Muzamil adds that the winners will sit where engineering meets the humanities: people who understand human behaviour and turn it into products step by step.

Taste is the new skill

Muzamil names a problem he sees often: “software slop.” People ask Claude for a product and get its most generic version. Good products need mindful decisions, like cutting a sign-up step or reducing 25 steps to a few. He asks how young people in Pakistan, with limited exposure to world-class products, can build taste.

Raheel agrees design is a major gap. Claude’s design is decent, he says, but people can tell it was made by AI. What helps is human psychology and observation. He made music and keeps a diverse circle. The idea for pakistangov.online came from a non-technical friend at 3 am on WhatsApp. His method for building taste is to follow great founders and creators and study what they read and notice. When he wanted to master Python, he studied what the creator of FastAPI was doing.

Muzamil adds a practical trick. Take a technical tweet you do not understand, paste it into Claude, and ask it to explain it “like I am five years old.” After ten or fifteen tweets, you see nobody is talking about rocket science.

System design before code

Raheel says he has barely written code by hand for a year and a half. “98 percent of my time is used in just thinking how to build and what to build,” he says, and whether it is worth building at all.

He gives the example from his own tool. Claude first added a Google sign-in. Raheel removed it, because a tool for a mass audience must be frictionless. That raised a new problem: how to protect it without a login, with thousands of bots scraping every site. That is where his technical background came in, with IP limits and cached queries. Left alone, he says, Claude will give you “a full-fledged Uber architecture” when you need a small app for four friends.

So he read five or six system design books in the last eighteen months, books he never read as a student. He tells Claude which patterns to use, debates the trade-offs and costs, and only lets it code once he knows exactly what he is building and for whom.

The opportunity inside Pakistan

Muzamil shares a story. A man in Peshawar who had left the army, with no technical background, watched one of Muzamil’s videos about building with Claude. He built an ERP for his gym, then his hair salon, then sold it by subscription to ten other owners he knew. Every technical person tells Muzamil there is no distribution. This man found it. Software has become so cheap that a business doing several million rupees in sales will happily pay a monthly fee if it makes the work easier.

But Muzamil also names the obstacle. AI needs good input data, and many Pakistani businesses still run on paper ledgers. You may have to build the base layer before you can sell intelligence on top.

Raheel agrees the market is untapped. A friend who runs a coffee shop built his own point-of-sale system and now shares it with other cafe owners. Investors, he says, now prefer AI-native companies built with domain experts, such as a law firm run on AI, over plain software. The money is in “boring businesses”: textiles, football and clothing manufacturing, Karachi’s industrial areas. These need people who go into operations, understand the problem, and build for it. He notes the world calls this a forward-deployed engineer. Pakistani startups used to fail by copying foreign consumer ideas into a market with little spending power. Now, he says, you can sell to “the people who actually have money.” His own new product is a voice-powered tool for real estate sales, and the response has been welcoming.

Four months in Southeast Asia

To close, Muzamil asks about Raheel’s travels. Muzamil keeps returning to the idea that exposure is powerful, and that Pakistanis resist it.

Raheel says travel was his retirement plan: a beach, a watermelon, and a laptop. Then he realised he could do it now. His friends kept saying no, so he booked a solo trip, his first time abroad. He later spent four months in Thailand, Laos, Cambodia, and Vietnam. He describes Chiang Mai as a kind of mini Silicon Valley for online workers, where coworking spaces and coffee are cheaper than in Pakistan. Most of the visas were online. When he needed a visa for Laos, he went to its embassy in Bangkok and had it within two hours.

Muzamil closes on that note. “If barriers are what you build in your mind,” he says, and you are “delusional enough,” you can make things happen. He hopes the government uses Raheel’s work, and the work of people like him, to modernise public services. To viewers, Muzamil says he is in the excited camp. Jobs will go for old, inefficient ways of working, but new opportunities are appearing, and those who make “a slightly uncomfortable pivot” now will have an early advantage.