A doctor who decided the better operation was on how we learn
Yermie Cohen runs Memorang, and Memorang has a deceptively simple promise printed across its homepage: capture your expertise, build in days, launch globally. Underneath that is something harder. Cohen is building what he calls an “AI stack for education” - the plumbing that lets a publisher, a university, or a professional association turn what they know into courses, quizzes, and exams without hiring an army of instructional designers. The pitch is not “here is another study app.” It is “here is the machinery to make a thousand of them.”
It is a strange place for a licensed physician to end up. Cohen trained to cut, to research surgical robots, to read anatomy slides at five in the morning. Instead he spends his days thinking about retention curves, content pipelines, and the uncomfortable question every educator is now asking out loud: how much of the teaching can a machine do, and how much should it.
His answer is unfashionably moderate. Let the AI generate. Keep the human grading. Memorang's whole architecture leans on that idea - automation for speed, oversight for trust.
He thought med school would be easy. It was the flashcards that broke him.
Cohen arrived at the David Geffen School of Medicine at UCLA with two undergraduate degrees from MIT - one in biology, one in mechanical engineering - and the quiet confidence of someone who had already survived the hardest engineering classes in the country. Medicine, he figured, would be a manageable next step.
It was not. The volume of material to memorize each week was relentless, and the tools on offer were a bad joke: either toy apps too simple to be useful, or platforms too expensive to be fair. An engineer's instinct kicked in. If the tool you need does not exist, you build it. With co-founder George Courtsunis, he started Memorang in 2013 as a free web and mobile study app that used spaced repetition, quiz modes, and games to drill information into long-term memory.
The first users were students like him - people in higher education and the health sciences who needed to learn enormous quantities of material and keep it. Memorang partnered with publishers, institutions, and authors so the flashcards were not just plentiful but trustworthy. The company appeared in NSF and MOOC research conversations as early as 2014, and by 2017 Cohen was on Rod's Pulse Podcast and in Inside Higher Ed explaining how health-sciences students could study smarter.
The frustration
Existing study tools were “either overly simplistic or prohibitively expensive.” So he made his own.
The method
Spaced repetition - showing you a fact right before you'd forget it - has been Memorang's spine since day one.
The partners
Publishers, universities, and professional bodies supply the content so the quizzes are trusted, not crowd-sourced guesswork.
From study app to the stack underneath the study apps
The interesting move was not building a flashcard tool. It was refusing to stay one. When generative AI arrived, plenty of edtech founders bolted a chatbot onto their product and called it a strategy. Cohen went the other direction and rebuilt Memorang as infrastructure - a platform that lets any organization launch customized, AI-driven learning across any topic, in days rather than quarters.
That repositioning came with grown-up requirements. Memorang completed a SOC 2 Type II examination, the kind of security and privacy bar that publishers and professional credentialing bodies demand before they hand over their content. The company's stated values - customer-centric, hustle, growth mindset, ownership, excellence - read like a manifesto for a team that wants to move fast without breaking the trust that test material lives and dies on.
Cohen now talks publicly about AI in test development and credentialing security, including on PSI's “Tried and Tested” podcast, where the recurring theme is the same one baked into his product: AI can generate exam content at a scale no human team could match, but humans have to stay in the loop to guard quality, security, and fairness. Speed without trust is worthless in a field where a single bad question can sink a certification.
Days, not quarters
Organizations capture expertise and stand up learning experiences fast.
Any subject
Content, quizzes, and assessment generated across topics, then deployed globally.
Human in the loop
SOC 2 Type II, plus human oversight on quality, security, and fairness.
Surgical robots, Navy desalination, then a quiz app
Before education claimed him, Cohen's research path wandered through some genuinely odd rooms. He worked with the Biomechatronics Group at the MIT Media Lab and inside the Harvard-MIT Health Sciences and Technology division. He consulted on mechanical engineering for a desalination project - a joint venture between UCLA, Sea Recovery, and the US Navy - figuring out how to pull fresh water from the ocean. At UCLA he researched at the Center for Advanced Surgical and Interventional Technology, managed a database lab at Wang Lab, and taught anatomy to medical students.
It is an unusually wide arc for someone who now sells education software, and it shows up in the product. Memorang is the work of a person comfortable with hardware constraints, biological systems, and the messy reality of how humans actually retain information. He had also already founded a company before this one - Enki, where he was co-founder and CEO - so Memorang was not a first rodeo.
Bars illustrate the span of Cohen's documented work history, not a formal ranking.