TLDRocket
Sign in

From leading products to leading with purpose: the 20-year product veteran building her own AI platform

Startups Magazine Sadia Hamid

A product leader in Sheffield built her own AI exam platform after two decades on other people’s roadmaps. WizardTeacher mixes test prep, university planning and UK-based data handling in one membership.

Based on reporting by Startups Magazine, Sadia Hamid — read the original for the full story.

Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error

Sadia Hamid spent 20 years making other people’s products work before deciding to build one with purpose at the centre. Her career ran through telecoms, ecommerce and edtech, and in Dubai she led digital transformation for a nine-million-customer base, re-platforming more than 10,000 products and putting machine-learning personalisation into production. Most recently, she took AI products from idea to production at a global edtech marketplace.

Then her last role ended earlier this year, and the usual job hunt did not go well. One rejection said she was overqualified. Another said she was underqualified. They landed in the same week. That was the break point. After two decades of asking for permission, she stopped.

The result is WizardTeacher, built for students preparing for tests such as IELTS, TOEFL, PTE, SAT and ACT, and for the rest of the journey after that exam. The pitch is simple: one membership instead of several subscriptions, with expert preparation priced for students rather than families that can pay corporate rates. It also tries to cover the wider path, from shortlisting universities to thinking about the career that comes next.

The most striking part is how hands-on Hamid has been. She says she designed and built the platform herself using Claude’s API, Stripe, Supabase and more than a dozen services, with the AI running in a London region so student data stays in the UK. A practice essay can come back in seconds, marked against official criteria with the exact sentences that lost marks and suggested replacements. That used to mean a trained examiner and a week of waiting.

Hamid is also unusually blunt about what AI can’t do on its own. It writes bad code with the same confidence as good code, she says, and what carried the product over the line was not the model but twenty years of testing discipline, data quality and knowing what not to build. WizardTeacher is due to launch its first phase in the coming weeks at wizardteacher.com.

My take — AI-written commentary, not fact-checked reporting

This is the healthy version of AI product building: less demo theatre, more judgement. The industry loves pretending the model is the hero, but the annoying truth is that the scarce skill is still knowing what users actually need and where not to trust the machine. The real novelty here is not that a platform was built with AI; it’s that someone used AI without handing over the steering wheel.

Read more about this at: Startups Magazine

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.