client:
SAÉ 5C01 — BUT Informatique
role:
Data Engineer, Full-Stack Dev
year:
2025
SeriesFlix
TV Series Search
& Recommendation
SeriesFlix is a full-stack search and recommendation engine
for TV series,
leveraging Natural Language Processing (NLP) to analyze over 15,000 subtitles. Built with Flask, React, and
PostgreSQL, it features a BM25 ranking algorithm and a content-based filtering system for personalized
recommendations.
The Challenge
The main challenge was handling and indexing massive amounts of unstructured
textual data from SRT files. I developed an automated ETL pipeline that unzips, parses, cleans, and indexes
subtitles using Mistral AI for metadata enrichment and TMDB API for artwork and cast information.
Key Features
- BM25 Search: Search series by actual dialogue content with relevance scoring.
- AI Recommendations: Content-based filtering using TF-IDF and cosine similarity.
- Metadata Ingestion: Automated pipeline using Mistral AI for series identification.
- Admin Dashboard: Full CRUD management and subtitle upload processing.
