Skip to main navigation menu Skip to main content Skip to site footer

Information technologies and systems

July 4, 2025; Zurich, Switzerland: VIII International Scientific and Practical Conference «GRUNDLAGEN DER MODERNEN WISSENSCHAFTLICHEN FORSCHUNG»


SCALABLE ARCHITECTURE AND AI-AUGMENTED DELIVERY IN ASTARLY LEARNING PLATFORM


DOI
https://doi.org/10.36074/logos-04.07.2025.022
Published
17.09.2025

Abstract

Astarly is built on an event-driven asynchronous architecture, structured around the user lifecycle: lead → payment → access → progress. The backend is implemented using FastAPI with PostgreSQL via SQLAlchemy ORM. Secure authentication is ensured through JWT tokens, two-factor authentication (2FA), and time-based one-time passwords (TOTP), including email verification. The AI component relies on an external API (OpenAI / DeepSeek) and internal moderation logic, using a risk-prefilter based on heuristics [1]. The decision-making process of the chat assistant is formalized through a probabilistic relevance model:

References

  1. Baker, B., Huizinga, J., Gao, L., Dou, Z., Guan, M. Y., Madry, A., Zaremba, W., Pachocki, J., & Farhi, D. (2025). Monitoring reasoning models for misbehavior and the risks of promoting obfuscation. arXiv. https://doi.org/10.48550/arXiv.2503.11926