Athena
Loading...
Date
2025
Authors
Tala AbuSoud
Raya Thawabe
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The podcast is among the strongest platforms for education, entertainment,
and information. With this aim, Athena has been developed
as a niche audio content platform to empower independent
creators and provide a richer listening experience for the audience.
Unlike the mainstream platforms, Athena caters to different user
needs with a blend of localized and niche-specific features.
Athena was developed via an iterative life cycle model approach,
planning, and initial research to understand user and market needs,
followed by UI/UX design, back-end and front-end development,
and appropriate testing. The system implements Flutter in the mobile
and web front-end, Node.js and PostgreSQL in the back-end,
and two Python-based FastAPI microservices for content recommendation
and audio fingerprint matching.
Further with real-time transcription supported by OpenAI Whisper
and recommendation engine powered by TF-IDF and cosine similarity,
audio fingerprinting is done by Dejavu, while a GPT chatbot
supports summary, translation, and QA functionalities. OpenAI
tools moderate user reviews; Cloudinary handles media streaming
and caching.
Two FastAPI-based microservices were developed: a recommender
system that uses user interactions, text similarity, and mood data,
and a matcher system that uses fingerprinting to identify podcasts
from short audio clips. Additional features include notifications,
JWT-secured authentication, an admin dashboard to explore app
and user insights, manage ads, and channel approvals, and a YouTube