Python, NumPy, Flask, BeautifulSoup

Sept 2024 - Present

While this was supposed to be a four-person group project, I thought I would challenge myself and program the entire search engine myself. In doing so, I taught myself the ins and outs of how search engines work: tokenization, cosine similarity, and massively optimizing the search to be efficient.

  • Built a scalable search engine indexing 55,000 URLs designed for corpus growth
  • Averaged <100ms latency for multi-token queries and <10ms latency for single tokens
  • Implemented TF-IDF score & cosine similarity to efficiently rank and score webpages
  • Efficiently handled memory usage to be less of the size of the corpus (~130MB)