Thought Behind Things

How a mechanical engineer built cricket AI from one camera

Saad of Sports Vector on tracking cricket balls with one camera, selling to federations, and taking padel analytics to 30 courts by year end.

  • Ep 551
  • Nov 5, 2026
  • 12 min read

Why a mechanical engineer is the right guest

The episode opens with Muzamil pushing back on a familiar complaint. Pakistani mechanical engineers often say they are finished, with no jobs and no work. Muzamil holds the opposite view. He argues that the old plan of picking a degree and waiting for the job it promises has to change. “You will essentially have to become a problem solver,” he says, and not even a worker. If you are a problem solver, you look left and right, pick up a stone, and find a problem underneath. In Asia especially, he says, there is an incredible number of problems to solve.

He adds that AI makes this cheaper, faster and more efficient. People who once said they had the vision but needed someone to fund them can now show a working demo for very little money. Muzamil frames the conversation with Saad around that idea. Saad is the founder and CEO of Sports Vector, and he has built a sports analytics company around it.

What Sports Vector actually sells

Asked for a 90-second pitch, Saad calls Sports Vector “a data analytics and strategy company.” It takes video data and live data and runs both through its machine learning systems and computer vision technology. It then converts the output for different buyers.

For broadcasters, it supplies intelligent overlays. Saad notes that five or six companies usually do that work, and Sports Vector does it as one. For franchises, it supports auctions, player dossiers and AI analytics. For federations, it helps with scouting and in-game training. In padel, it installs cameras at courts so that players get analytics.

Saad says the company has worked for about a year and a half. He says it has a few clients, has moved past the demo stage and has worked with franchises and international teams. By his account, it has run sessions with five to six coaches.

From public policy to 40 products

Muzamil asks about Saad’s background, and the answer is unexpected. Saad graduated in 2024 and says he had no interest in entrepreneurship at all. His plan was to study, return to Pakistan and go into politics, and he says he was targeting Taimur Jhagra’s kind of path. He then did a master’s in public policy.

During that degree he came to believe that, to have impact, entrepreneurship could beat the public sector. That is how he got into AI. His account of the learning curve is blunt: “I have built probably personally 40 products in the last two years.” Sports Vector is the one that became the main focus. Without the other 40 and the difficulties along the way, he says, he would not be where he is.

Demos are the apprenticeship

Muzamil connects this to a criticism he hears from older people in software and tech. They say AI only lets young people make demos. His answer draws on his own start in media, which he says he entered as an outsider with no business being there. When the costs of production, post-production and distribution collapsed, people like him came in and built a business model through exploratory content. He sees the same pattern in software now.

His claim is direct: the first ten vibe-coded projects “are gonna be demos and they are gonna be useless.” But fifty demos lead to a product, because it is learning at scale and getting your hands dirty instead of sitting through a course or a degree. His advice is to build stupid demo after stupid demo, because that is how you find what works.

Saad agrees, and adds a limit. Production applications bring a checklist, with cybersecurity first and scalability second. There are parts of the product, he says, that he could not have built with AI had the knowledge not been there. He credits his co-founder Hamza for leading the computer vision: if Hamza had not understood how the model works, AI could not have built it. His analogy is that AI can act as the technician, but the intelligence layers need background and knowledge.

A bootstrapped team built at night

Sports Vector has two co-founders, Saad and Hamza. Saad says that when they registered the company last year they had no money and could not really hire anyone. Neither of them was a developer. Hamza knew machine learning and computer vision. Saad describes himself as a mechanical engineer and a very quick learner.

So he began hiring freelancers, one or two at first. He says he would sit with them from midnight to 4 a.m. and watch how they worked. He learned the work and also learned how to work with such people. Earlier, during four months of unemployment, he had written two books.

The grant changed the scale. Saad says it was 154,000 USD, received in December. After that he held around 200 to 250 interviews. The company now has 10 people, including five developers and two or three people covering marketing and sales. He says they also went looking for mentors and advisers. They spoke to high-profile people, but some asked for what he calls an insane percentage, so they gave equity to a group of people who help with networks, access and what to build. He adds that there have been many pivots since the start.

One camera, a Babar Azam plan and the delivery data problem

Muzamil asks Saad to show the product, and the cricket engine is the first demo. Saad says that only his company and one other do ball tracking for broadcast cricket, and the other has existed for more than ten years. He says his is the only Asian and only Pakistani company doing it. Output arrives with a latency of a few seconds, with length, line, deviation and bounce for each ball.

The motivation is domestic cricket. Pakistan has leagues and junior cricket, and Saad says there is “so much data that needs to be captured” but it is all done by hand today. A delivery takes two to three seconds, and a person must label eight or nine things in it. Any analyst, he says, will tell you the accuracy is poor. His answer is computer vision, and the point he stresses is that “we do everything through one camera,” not six.

On the platform, Babar Azam is rated against other international players in the powerplay and the middle overs, scored out of 10. Saad says that “anyone who played FIFA would know what those numbers mean.” The platform also shows performance over a batter’s first 15 balls and a wagon wheel. Clicking an area shows the video behind it.

The second pitch is that everything lives in one ecosystem. A team does not need to download data, put it in Excel and then paste it into Claude or GPT to build systems. You choose the players, search for one, and for a batting plan against Babar Azam you press generate. Saad says the plan builds automatically, so you have “a match plan against Babar Azam within a click of a button.” He says he has built player dossiers like this for one or two teams.

Why Sports Vector sells to organisations, not fans

Muzamil asks whether selling only to broadcasters and leagues caps the business, and whether Saad has considered mass consumers. Saad says the company first tried a business-to-consumer approach, then changed direction. His reason is price: no coach in Pakistan will pay 20 dollars a month unless an academy is paying.

So Sports Vector now sells directly to federations, associations and broadcasters. Saad names two benefits. The sales pipeline is smaller, so it saves time and energy, and each sale is worth more. Second, the market is large. By the company’s analysis, analytics alone, without broadcasting, is a 25 to 30 million dollar market in USD, and he says Sports Vector has captured maybe 1 percent or less. He says he would be “more than happy to be limited at 30 million” in cricket.

He says the plan is to go to padel next, which he calls the fastest-growing sport in the world. He says the company has sold more in padel in two months than in cricket in eight. He is asked why not football, since more people play it. His reply is that in the GCC, and in Qatar specifically, cricket is actually the most widely played sport, and it is also fast-scoring.

Padel, revenue share and the web app

Muzamil raises a parallel. He found another founder on LinkedIn who is building experience centres in Pakistan, and he describes a US company called Toca. Muzamil says it runs indoor sports for private academies and uses white-labelled Chinese hardware with cameras, real-time analytics and screens, and he says it is valued at half a billion dollars. His reading is that amateurs want analytics too, and he sees a clear opportunity in volume.

Saad describes the padel model. When he spoke to a Karachi court, they told him their commercial model was very different. Courts that already work with an AI company pay a cost per court, and they were unwilling to pay one more. So Sports Vector proposes revenue sharing. It integrates with the court’s booking platform, or it installs a QR code at the court. You scan the code, the camera switches on, you play, and then you pay for the analytics.

There is no separate app, because Saad says he is “tired of downloading apps.” Players use a web app that works on mobile. After a game you select which players you were, and a player report costs between 850 and 2,000 per person depending on the location, which he says includes Doha and Pakistan. Courts can also place their logo on shareable content, which Saad calls a growth-hacking tool.

The differentiation is depth. Saad says most systems give shots, heat maps and speed. He wants answers like where Saad moved when Muzamil played a lob. He also wants a broadcast layer for padel that borrows from the F1 fan experience, with heat maps, skeletons and performance from past games. Auto-scoring is in the pipeline but not live, and the company currently integrates existing scoring APIs. The target is 30 padel courts by the end of the year. Saad says there are about five courts now, mostly in Doha and Pakistan.

A race to cost, and the work that remains

Later in the discussion, Muzamil pushes on a worry. If the idea itself is not the edge, “the differentiator will be the optimization around that idea and the actual execution.” In his view it will be a race to the cost. If one hour of processing costs 20 dollars and a rival gets it to 5, he says, the rival wins at scale. He also suspects Google credits are currently subsidising the business.

Saad concedes the credits, since the company has a large amount of Google and RunPod credits, but he gives two answers. Compute costs fall over time, and many parallel companies also run on the cloud instead of local GPUs. The second answer is optimisation: models improve with more data and can be quantised, so they run faster, and he says cloud pricing is per second or per hour. His model runs 10 to 15 minutes today and could drop to 8 or 9. Margins are high now, he says, and there are ways around them shrinking.

Saad then points to the harder work. Building is easier now, he says, but deep tech is still difficult and the harder part is business-to-business partnerships and building a trust layer. He says he visited the padel court five or six times in a month. He also had to find a subcontractor to install cameras. His summary: “100% I think distribution and ecosystem building is very important.”

The company has signed with a Qatar padel court that he calls one of the top three in the country. He says he is talking to another company with 600 cameras installed worldwide that wants to use the technology. He says inbound is always better than outbound in sales, and that the plan is to work with five or six courts first so the technology holds up.

Runway, fundraising and what Saad is not dreaming about

Muzamil asks about the health of the startup, with a six-figure grant, ten employees and two founders on pay. Saad says he cannot share too much publicly about fundraising. He does say the company has runway until March or April next year and that fundraising is going well. He says the company is working out which deal to take, and that once done it should reach 2028. He rates the company a B minus for now. He says it skips the pre-seed layer, arguing that pre-seed hardly makes sense in an age of AI, and the company is raising a seed round. He says it recently graduated from NIC Karachi, where revenue figures are publicly available, and that revenue is decent for a startup.

By the end of the conversation, Muzamil says he is in a phase of wanting to show the country that a new world has been created. Saad answers with restraint. He says this is not the time to dream about unicorns or becoming a billionaire. He calls it “a concept of tawakkul”: you put in the effort and keep going. Muzamil adds that the old obstacles of privilege, connections and knowledge have largely started to disappear. Five years ago, he says, nobody would have noticed this work.

Muzamil closes with a note of feedback for Saad. He tried five times to find the CricTier website and could not, and says the SEO is poor. He says he will put Sports Vector’s site in the description, and he invites viewers to try the product and tell Saad what they think. If they are in Karachi, he says, they can go to a padel court and get their analytics.

Build with me in the Inner Circle.

You don't need to code. You need a problem you know well, and a room of people building beside you.

Muzamil Hasan speaking on stage