Step 1: Build the Foundation We built the full interface and product system from scratch — everything beyond the MVP. That included: AI-powered tutoring insights (pre-LLMs): Automatically analyzed sessions to detect topics, sentiment, and confusion signals. Feedback loops: Tutor reports that were summarized and sent to teachers. Knowledge gap detection: Built dashboards to show teachers which students struggled with which concepts — and which topics they depended on. Scheduling tools: Tutor calendar system to match demand with availability. Demand forecasting: Built a prediction system with 90% accuracy to preempt surges and dips. Internal tools: Dynamic product roadmap system that scored features by impact and complexity. Feedback categorization system: Turned raw user feedback into usable, prioritized product insights. Interface design + branding: Led redesign of the product UI and full visual system. Documentation: Created team-ready slides, specs, systems, and pitch materials. This version of the product helped raise the $2M seed round. I worked closely with the CEO and CTO, contributing to both product and fundraising — from deck creation to investor meetings.
Step 2: Fix What Blocks Progress While scaling that product, I reviewed the session start flow. Something felt off. Students landed on a plain white screen with a box prompting them to select a topic. It worked, but it didn’t excite. Students hesitated. So I tried something simple: I overlaid that same topic selector on top of the actual chat interface. Just a peek. A preview of what was underneath. Now they saw what came next — and it clicked. Session starts went up by ~25%.
Step 3: Build What They’re Asking For As more usage came in, a pattern emerged: “Can someone help me review my paper?” Tutors were already doing it — manually — in chat. But there was no dedicated interface. No system. So we built one. The Paper Review feature made it easy for students to submit work, get structured feedback, and for teachers to track it all transparently. This saved teachers 30%+ of their time — without losing the human connection. Now, they could log in, see all their students at a glance, spot the ones struggling, and review tutor notes and AI-generated insights. It gave them back control. For overloaded classrooms, this was a breakthrough. The company rebranded from GradeSlam to Paper — because this was now the core value.
Step 4: Keep Listening (and Translating) What made this all work wasn’t just one feature. It was the loop: Talk to users constantly (students, teachers, districts). Watch usage behavior. Listen to feedback. Categorize input by frequency and friction. Score ideas by impact vs. effort. Make changes fast. Check with the same users: “Does this help?” We didn’t just collect feedback — we built systems to turn it into product decisions. This is what let us move quickly and stay right.