Shashi

WitRush ↗

Web App

A semantic word game that spread entirely through word-of-mouth - no ads, no paid channels. 26k users found it organically in 50 days across the US, India, UK, and Australia. Sessions average 10 minutes - rare retention for a casual word game.

Zelyft ↗

Android App

A minimalist Android launcher that found its audience without a single dollar spent on marketing. 1,000+ users downloaded it because it solves a real problem, and a 4.9★ rating keeps it surfacing in Play Store search. Reviews consistently call it the cleanest launcher they've tried.

This project focuses on predicting the prices of used cars by building an end-to-end data pipeline and machine learning model trained on real-world listings scraped from Indian online marketplaces. The workflow includes web scraping, data preprocessing, feature engineering, and small neural network model training (TensorFlow).

This dataset was created to address a practical gap: there is no large, clean, India-specific used-car dataset suitable for machine learning price prediction. Public datasets are typically US-centric, sparsely populated, or filled with missing values. To solve this, a multi-source data collection pipeline was built to extract structured listing metadata directly from major Indian automotive marketplaces. The result is a 57,480-row, fully unique, zero-null, ML-ready dataset covering a wide geographic and brand spectrum across India.