Mikhail Shifrin Email & Phone Number

Уилмингтон, Делавэр, Соединенные Штаты Америки

Editing a book on AI and prompts for the user support

Висконсин, Соединенные Штаты Америки

As a Prompt Engineer utilizing GPT API, I achieved the following:* Analyzed thousands of student-teacher messages in Discord, synthesizing them into a comprehensive FAQ.* Developed prompts for an AI-powered automated assistant that aids students in programming tasks on the learning platform. This included generating hints in chat without revealing the correct solutions, successfully implemented and operational on the platform.* Created an automatic A/B testing system for new prompt revisions, which reliably compares the output of different prompts. A script was developed that accurately identifies the more effective prompt.* Addressed a major issue in AI-powered assistants - the unintelligibility of their hints - by rapidly refining prompts with A/B testing, eliminating the need for manual verification.

Inventing and making educative features. Analysis and assessment of educative technologies.I identified problems our bootcamp faced in the American market, proposed solutions to these issues, and was responsible for implementing these solutions.A significant problem was dozens of students who purchased the course but overestimated their own capabilities. They dropped out within the first two months, wasting their time and consuming the resources of our guidance team. By correlating each student's performance with data from in-depth interviews, we determined that students initially fall into four categories:1. Those who definitely cannot cope with the studies and quickly drop out.2. Those who will find it very difficult but can manage at least half of the course if they have more than 20 hours of free time per week.3. Those who will find it challenging but can succeed with active support from our side.4. Those whose experience is close to IT professions and who can independently complete the course - these students only drop out if they suddenly find a good job.I developed a student segmentation based on their background, life circumstances, and motivation, using 12 criteria. This resulted in a short entry questionnaire with checkboxes, which is quick and easy to complete during registration.Thanks to this segmentation, we could distinguish, even before the classes began, those who could not study from those who could and should be helped.Interestingly, this segmentation accurately predicts a student's behavior in the first two months of the bootcamp. After this period, students understand what they are dealing with, assess their success, and their motivation significantly changes. By that time, the support team has had a chance to get to know the students and develop a personalized approach to them.

Moscow, Russian Federation

Editing Yandex' online courses & coding tutorials.

Москва, Россия

https://www.facebook.com/DoktorEpic/