Prim (Maymani) A.

Projects

This is open list of my projects, sorted by recency. Currently working towards making the list longer...


Nonsensical Metaethics Paper Generator | Github | Demo (in-progress...)
creative coding, typst, metaethics, silly
  • A common quibble in a metaethics course I took was that the papers we read were too jargon-heavy or syntactically atrocious. (This complaint become even more pronounced during our week of reading Allan Gibbard.)
  • It inspired me to make this deterministic metaethics paper generator in Typst.
  • I think making this was one of the rare times where I was smiling the entire time of coding. It is silly to admit, but I had to catch myself from bursting into giggles from all the outputs.


Met Museum Self-Tour Curator | Github | Presentation Demo (Youtube) | UX Design
machine learning, recommmendation system, pytorch, natural language, embeddings
  • Target User: Met Museum visitors, The Metropolitan Museum of Art

  • Purpose:
    - Assist museum visitors in finding artworks grouped by a common theme of their choosing
    - We built this tool with the concept of structured flexibility in mind. We wanted to balance the visitor's sense of self-directed curiosity, while still providing enough scaffolding for a structured and memorable museum experience.

  • What is it:
    - A search/recommendation system over the Metropolitan Museum of Art's on-view collection (~44k artworks).
    - Input: takes a user search-query in English (ex. "melancholic landscapes").
    - Process: transforms user-query by 1. embedding the query, 2. probabilistic calculation of query alignment to artwork clusters, 3. use cosine similarity to selected relevant art in art clusters
    - Returns: a set of relevant art objects grouped by that query + an optimized walking tour-route in the museum for that user to take.


PulseCircle | Demo | Github | DevPost
healthtech, ux design
2nd Place Hackathon Winner – PulseFoundry x AI Healthcare
  • Developed proof-of-concept home cardiac rehabilitation app for healthcare hackathon enabling hybrid remote/in-person recovery plans, guided exercise sessions, AI patient support, and real-time care team monitoring.


RateMyProfessor (Teacher Review Platform Analysis) | Github
data/user analysis, eda, python, pandas, sci-kit learn, matplotlib
  • Analyzed ~90K user reviews from education review platform to identify drivers of student satisfaction and learning engagement.
  • Conducted Pearson and Spearman correlation analysis and PCA to examine relationships among ratings, difficulty, engagement signals, and course attributes. Visualized key-trends with Matplotlib.
  • Developed a logistic regression classifier (AUC = 0.76) to extract behavioral patterns from ~90K unstructured user reviews. Applied LASSO regression to identify statistically significant language-driven predictors of engagement.