Introduction
Samira Kühn wrote her first tutor prompt on a library computer because her own laptop had stopped charging. Around her, students packed away expensive devices. She stayed until closing time and tested whether the system could explain fractions without giving the answer away too quickly.
Samira lives in Cologne and studies while building an AI tutor for pupils who cannot simply buy private lessons. She is the first in her family to attend university, and that still shapes how she designs learning tools.
Story of the Path into AI
In school Samira had learned how much tutoring depends on money, quiet rooms and adults who know the system. In her first semester she wanted a tutor that explains thinking paths, not just solutions. The first versions were too eager. The bot praised wrong steps, jumped to formulas and sometimes made a child feel as if the answer should have been obvious.
Samira learned from open courses, forums and tests with younger siblings and neighbourhood children. She also learned that her own insecurity was not a private flaw but a useful test case. When a learner asked a so-called simple question three times, the system had to remain patient. It also had to admit when it was uncertain.
Her first stable project was a maths chatbot that breaks exercises into small steps and asks after each explanation what the learner understood. It took several awkward trials before it stopped sounding like a textbook with a smile.
Current Work
Today Samira works in a university project on fair educational AI. She collects feedback from schools in disadvantaged neighbourhoods and rewrites the tutor around real questions. In one test the bot solved word problems too quickly because it recognized patterns. Samira added a step that checks vocabulary and prior knowledge first.
The effect is not a miracle. Some pupils say they dare to ask the tutor “stupid questions” before speaking in class. That matters to Samira. The goal is not to remove teachers, but to reduce the shame that stops questions from being asked at all.
Personal Advice
“Use your uncertainty as a test case,” Samira says. If you felt lost in a classroom, remember the details: where the explanation moved too fast, where the language became heavier than the task. Those details are design material.
Key Facts
Age and place: 20, Cologne.
Background: first-generation student, side jobs, limited equipment.
Entry into AI: maths chatbot with step-by-step explanations and follow-up questions.
Focus today: educational AI.
Typical tools: language models, learning analytics, simple web development.
Werkstattnotiz
Samira keeps the prompt in which the tutor congratulated a wrong answer. It sounded kind and did damage. She now checks praise as carefully as correction. Her next test concerns silence: how long should a system wait before helping without making the learner feel watched?