Voxza

The Voxza was developed to address the core technical problem was that video upload and ai analysis were too heavy to run inline in the fastapi request path: ffmpeg audio extraction, google speech transcription, and gemini summarization/insight generation create long latency, high memory usage, and fragile failure handling when the user waits for the upload to finish This project showcases my skills in Python, FastAPI, SQLAlchemy, and other technologies.

During development, I focused on creating a solution that would be both effective and user-friendly. The approach involved the project separated the fast path from the expensive path by moving analysis into a celery + redis background pipeline. the api accepts uploads quickly, stores metadata in postgresql/supabase, and queues processing jobs; workers then stream video from gcs, extract audio in parallel with ffmpeg, transcribe with google cloud speech, and analyze content with gemini, updating result state asynchronously for processed or failed jobs.

The application features a clean, intuitive interface that allows users to easily navigate and utilize all functionality. Key features include Short-video capture and upload, AI transcription and content analysis, Emotion, mood, and action-item extraction, and more.

This project demonstrates my ability to understand complex problems and develop effective solutions using modern technologies and best practices in software development.