Work
Jesa AI
Industry: EduTech / AI
Deliverables: Product Strategy, Design & Launch
Year: 2025
Jesa AI — Winner, Kelley AI PM Hackathon 2025
Current tools like ChatGPT, NotebookLM, and Quizlet each solved one piece of the puzzle but none of them adapted to student performance, generated multilingual video content, or grounded their outputs in the student's actual course documents.
Demo Video of Jesa AI
Snapshot of AI Architecture of Jesa AI
Winning the Kelley AI PM Hackathon was a full validation of the approach, from how we researched the problem to how we designed the solution. But beyond the win, building Jesa AI sharpened how I think about AI product strategy in ways I carry into every project since.
What I learned
Responsible AI should always be a core product decision, not an afterthought. Privacy-first design, hallucination prevention through RAG, and FERPA compliance were built into the architecture from day one.
A tight problem scope wins. We did not try to replace the entire learning experience. We identified the exact moment students hit a wall and built precisely for that moment. That specificity is what won the hackathon.
User research drives technical decisions. Every guardrail, every model choice, every UX surface had a direct line back to something a real student told us in an interview.
Trade-offs are strategy. Choosing GPT-4 over cheaper models, capping video generations, using RAG over general prompting, each decision was a deliberate strategic call balancing cost, accuracy, and user value.
(6)
