AI Tutor

Learning platform with an AI pipeline including deep knowledge diagnostics, dynamic theory tailored to each student, and assignment help.

Service features

• Knowledge decomposition into individual skills with theory, practice, and assessment tasks
• Skills graph for visualizing connections between learning topics
• Prerequisite system for gradual topic unlock
• Deep knowledge diagnostics using a Knowledge Tracing model
• AI-generated dynamic theory on demand for each student
• Skill practice with an AI assistant and automatic assignment checking

Key achievements

• On the AI Prototypes team, developed educational AI services to validate EdTech hypotheses; the solution reached production integration
• Initiated a new theory model and built a skills graph with React Flow — simplified topic navigation and learning path adaptation
• Set up Next.js ISR and manual theory page invalidation: LCP dropped from 3.9s to 1.1s, Speed Index from 2.4s to 0.7s; learning content became indexable
• Replaced manual theory authoring in a custom MD format with a TipTap WYSIWYG admin panel and custom blocks, speeding up content layout for instructional designers
• Moved TikZ-to-SVG rendering to a dedicated NestJS service with Redis caching — avoided adding 5 GB of LaTeX dependencies to the main backend
• Set up OpenAPI code generation with Kubb/Orval, reducing manual API contract maintenance