Nidhi Mendre

I’m a self-driven data science enthusiast taking an unconventional path to build hands-on, technical, and entrepreneurial projects. I focus on applying data-driven approaches to solve practical problems and create measurable business value.

In my projects, I forecast Uber ride demand using geospatial cleaning, Haversine-based features, and MiniBatchKMeans clustering, training regression models that reach ~90% R². I also built an end-to-end Swiggy delivery time prediction pipeline with feature engineering and Optuna-tuned models—LightGBM delivered MAE = 3.06 and ~35% improvement over baseline.

I’ve developed a hybrid Spotify recommender system using content-based and collaborative filtering, analyzing 120k+ tracks and 15+ audio features. By combining cosine similarity with matrix factorization, I achieved ~92% recommendation accuracy and improved personalization by 40%.

Through internships, I’ve worked on automation and R&D—building a Selenium-based WhatsApp automation flow and supporting research, social media campaigns, and decision-making with Excel models and presentations. I’m eager to contribute to impactful data science initiatives with real-world outcomes.