This is a guest post by Sahin Boydas, a serial entrepreneur and angel investor. He has built and sold two companies, RemoteTeam (acquired by Gusto) and MovieLaLa (acquired by Gfycat), and has invested in over 200 startups, including Anthropic, OpenAI, and Scale AI.
Fifty thousand students. That’s not a vanity metric. That’s 50,000 individual learning journeys, 50,000 different paces, 50,000 unique sets of strengths and weaknesses. When we first started building our personalized learning platform, I thought I knew what I was getting into. I had the MBA, the tech background, the "move fast and break things" mentality. I was wrong. Scaling an AI learning platform to that many users taught me more than any degree ever could. Here are the seven biggest lessons I learned from the trenches.
1. Personalization is More Than Just a Name
I’ve seen so many so-called “personalized” learning tools that do little more than insert a student’s name into a generic worksheet. That’s not personalization. That’s a mail merge. True personalization is about understanding how each student learns and adapting the content and delivery to their individual needs. We spent months building a system that could analyze a student’s performance in real-time and adjust the difficulty of the material on the fly. It was a massive undertaking, but it was the only way to create a truly effective learning experience.
We used a combination of collaborative filtering and content-based filtering to make our recommendations. Collaborative filtering looked at the behavior of similar students to predict what a new student might like. Content-based filtering analyzed the content of the learning materials themselves to find items with similar attributes. It was a complex system, but it was worth it. We saw a 30% increase in student engagement after we rolled it out.
I remember one student in particular. Let’s call her Sarah. She was struggling with algebra. Our system detected that she was having trouble with a specific concept – solving for variables. Instead of just giving her more of the same problems, it served her a video tutorial that explained the concept in a different way. Then it gave her a few practice problems, and when she got them right, it moved her on to the next concept. That’s the power of true personalization.
2. Engagement is Everything
You can have the most sophisticated AI algorithm in the world, but if students aren’t engaged, it’s worthless. We learned this the hard way. We launched with a beautiful, minimalist interface that we thought was the epitome of good design. The students hated it. It was boring. We had to go back to the drawing board and add gamification elements, like points, badges, and leaderboards. It felt a little silly to me at first, but it worked. Engagement shot up, and so did learning outcomes.
We ran A/B tests on everything. We tested different point systems, different badge designs, and different leaderboard configurations. We found that students were more motivated by intrinsic rewards, like seeing their progress and mastering new skills, than by extrinsic rewards, like virtual currency. We also found that social elements, like being able to see their friends’ progress, were a huge motivator.
One of our most successful features was a “streak” counter that showed students how many days in a row they had used the platform. It was a simple idea, but it was incredibly effective. We had students who would log in every single day, just to keep their streak alive. It was a powerful example of how a small design change can have a big impact on user behavior.
3. The Ethical Tightrope of AI in Education
When you’re dealing with student data, you’re walking an ethical tightrope. You have a responsibility to protect their privacy and ensure that your algorithms are fair and unbiased. We had a lot of sleepless nights worrying about this. We ended up building a system that gave students complete control over their data. They could see what we were collecting, how we were using it, and they could delete it at any time. It was a huge engineering challenge, but it was the right thing to do.
We also had to be very careful about algorithmic bias. We knew that our algorithms were only as good as the data they were trained on. If our training data was biased, our algorithms would be biased too. We spent a lot of time and effort cleaning our data and making sure that it was representative of our diverse user base. We also built in a system for detecting and mitigating bias in our algorithms. It wasn’t perfect, but it was a start.
I’ll never forget the time we discovered that our algorithm was recommending easier content to students from low-income schools. It wasn’t intentional, but it was happening. We had to go back and retrain our algorithm with a more balanced dataset. It was a wake-up call for us. It made us realize that we had a responsibility to be proactive about fighting bias, not just reactive.
4. Data is Your Most Valuable Asset (and Liability)
In a personalized learning platform, data is everything. It’s how you understand your users, how you improve your product, and how you measure your impact. But it’s also a huge liability. A data breach could be catastrophic, not just for the company, but for the students whose data was compromised. We invested heavily in security and compliance, and we made sure that we were only collecting the data we absolutely needed.
We encrypted all of our data, both in transit and at rest. We implemented strict access controls, so that only authorized employees could access student data. And we had a third-party security firm audit our systems on a regular basis. It was expensive, but it was worth it for the peace of mind.
We also had a very clear data retention policy. We only kept data for as long as we needed it to provide our service. And we made it easy for students to delete their data at any time. We wanted to be as transparent as possible about how we were using data, and we wanted to give students as much control as possible over their own information.
5. Teachers are Your Partners, Not Your Customers
We made a big mistake early on by trying to sell our platform directly to schools. We thought that if we could just convince the administrators, the teachers would follow. We were wrong. Teachers are the ones who are in the classroom every day, and they know what their students need. We had to change our approach and start working with teachers as partners. We listened to their feedback, we incorporated their ideas, and we built a product that they actually wanted to use.
We created a teacher advisory board, made up of some of our most active and engaged teachers. We met with them on a regular basis to get their feedback on new features and ideas. We also created a community forum where teachers could share best practices and learn from each other. It was a lot of work, but it was worth it. Our teacher community became one of our biggest assets.
I remember one teacher who told us that our platform had completely changed the way she taught. She said that it had freed her up to spend more time working one-on-one with students who needed extra help. That’s the kind of impact we wanted to have. We weren’t just building a product. We were building a tool that could help teachers be more effective.
6. The "Cold Start" Problem is Real
One of the biggest challenges in personalized learning is the “cold start” problem. When a new student signs up, you have no data on them. How do you personalize their experience? We tried a bunch of different things, from diagnostic quizzes to user-selected learning paths. In the end, the best solution was the simplest: we just asked them what they wanted to learn. It sounds obvious, but it’s amazing how many companies forget to just talk to their users.
We created a simple onboarding flow that asked students to select their interests and goals. We used that information to create a personalized learning plan for them. It wasn’t perfect, but it was a lot better than starting from scratch. And as students used the platform, we collected more data and were able to refine their learning plan over time.
We also found that it was important to give students a sense of progress right from the start. We created a “starter pack” of content that was easy to complete and gave students a quick win. It was a small thing, but it made a big difference in getting students to stick with the platform.
7. Simplicity Scales, Complexity Fails
As we grew, there was a constant temptation to add more features. We had a million ideas for new things we could build. But we quickly learned that simplicity is the key to scale. Every new feature adds complexity to the product, and complexity is the enemy of a good user experience. We had to be ruthless about prioritizing and only building the things that would have the biggest impact on our users. It was a painful process, but it was essential to our success.
We used the “KISS” principle – Keep It Simple, Stupid. We had a one-in, one-out policy for new features. If we wanted to add a new feature, we had to get rid of an old one. It forced us to be really disciplined about what we built. It also forced us to focus on the core value proposition of our product.
I remember one time when our product team came to me with a proposal for a new feature that would allow students to create their own avatars. It was a cool idea, but it had nothing to do with our core mission of helping students learn. I had to say no. It was a tough decision, but it was the right one. We had to stay focused on what really mattered.
Building a personalized learning platform for 50,000 students was the hardest thing I’ve ever done. It was also the most rewarding. I learned more about technology, about business, and about myself than I ever thought possible. And I got to see firsthand the power of AI to transform education. The future of learning is not about replacing teachers with robots. It’s about empowering teachers with tools that can help them reach every student, one personalized learning journey at a time.
Frequently Asked Questions
Can I implement all of these at once?
I'd strongly recommend against it. Pick the 2-3 items that resonate most with your current situation and focus there. Trying to do everything simultaneously is a recipe for doing nothing well.
How were these items selected?
Each item on this list comes from direct experience, either from building my own companies or from patterns I've observed across the 200+ startups I've invested in. I prioritize practical, actionable items over theoretical concepts.
How do I know which items apply to my situation?
Start by honestly assessing where your biggest bottleneck is right now. The items that address that specific constraint will give you the highest return on your time and energy.
Which item on this list has the highest impact?
It depends on your stage and context, but in my experience, the items near the top of the list tend to have the broadest applicability. That said, sometimes the less obvious items create the biggest breakthroughs for specific situations.