80
%
questions resolved by the bot
80% of student questions are handled automatically; live curators only see complex cases.
About the Project
We built and deployed an intelligent curator bot that fully handles student support, progress tracking, and access synchronization for a large crypto trading online school.
This is not a simple auto-responder — it is a full-fledged digital employee integrated into the school's business processes.
How It Works for Students
For users, the bot acts like an attentive, always-available mentor.
24/7 support
Students can ask about lessons, terminology (e.g., “what is shorting?”), or technical details at 3 AM and instantly receive a detailed answer with links to exact course materials.
Feels human
The bot knows the schedule of live curators (Vladislav, Marina, Pavel) and adapts its tone to whoever is on duty. If Marina is on shift, the bot speaks in a feminine voice, matching her expressions and emoji style.
Example of a student's Q&A with the bot
GetCourse reminders
Automations within GetCourse track every student’s course progress. The bot reminds them to submit homework, congratulates them on finishing modules, praises completed lessons, and offers help with specific topics.
Reporting workflow
Once a week the bot gently nudges students to submit trading stats. Inside Telegram, users fill an interactive survey, attach screenshots, and the bot collects everything neatly.
Personalization
The bot addresses learners by name, remembers context, and even handles courteous small talk—building a sense of care rather than talking to a machine.
Business impact
For the client company, the solution covers several critical tasks at once:
Lower payroll load
The bot closes up to 80% of routine requests so live curators focus on complex cases and VIP clients.
Automated access control
- Integrated with GetCourse: if payment stops or access expires, the bot checks status via API and politely limits premium features.
- When a course is purchased, the student is onboarded instantly—no manual admin actions.
Centralized analytics
All performance stats (profit, PnL, trade counts) automatically sync to Google Sheets and the internal DB. The school sees real-time effectiveness of its trading strategy in students’ hands.
Better retention
Scheduled reminders keep students engaged, boosting course completion (COR).
Tech stack & architecture
IT-AI built the system on a reliable, scalable stack with a focus on security and response quality.
Core & AI
Language
Python 3
Framework
python-telegram-bot (asyncio)
LLM
OpenAI GPT-4o / GPT-5
RAG
LangChain + FAISS
RAG knowledge base: LangChain + FAISS store all lessons, video transcripts, and FAQ. Before answering, the bot fetches relevant chunks to eliminate hallucinations and ensure methodological accuracy.
Intent classification: A blend of heuristics and AI classifiers detects whether it’s a course question, small talk, or a prompt-injection attempt.
Integrations and Data
- GetCourse Webhooks: Our custom Flask server receives webhooks from GetCourse (registration, payment, status changes) and instantly updates user permissions in the bot.
- Google Sheets API: Seamless export of student surveys and reports to tables for the analytics department.
- PostgreSQL: Reliable storage of user profiles, access history, and logs. Implemented connection pooling (psycopg2 pool) for stable performance under load.
- Bizon365: Integration for synchronizing webinar data.
Security and Reliability
Fail-safety
Implemented global error handlers and automatic restart systems. Even if one module fails, the bot continues to respond to users.
Protection
Flexible message filtering system, spam protection, and rate limiting.
Logging
Detailed rotating logs of all actions for quick debugging and auditing.
Results of Implementation
24
/7
support availability
The bot is available 24/7, providing instant answers to questions at any time.
100
%
access automation
Full integration with GetCourse for automatic access management.