Devansh
Tyagi.

Exploratory data analysis and time-series visualization of a decade of NASA Giovanni satellite data to track Delhi-NCR's shifting pollutant landscape.

• Improved data consistency by 70% through advanced preprocessing on large-scale multi-year environmental datasets.
• Identified a 15% decrease in CO and 20% increase in NOx emissions via time-series forecasting across 10 years of satellite observations.
• Produced geospatial visualizations of pollutant distributions over the Delhi-NCR region using GeoPandas and Matplotlib.

Full-stack movie recommendation platform powered by a custom ML model serving personalized picks in real time.

• Built a custom ML recommendation model improving content relevance and personalization by 30%.
• Deployed RESTful APIs via FastAPI delivering real-time recommendations with sub-second response times.

Minimalist always-on-top Windows utility that pulses a single dot when a chosen WhatsApp contact messages you — no notifications, no popups.

• Built a frameless, transparent, click-through Electron overlay that floats above all windows including fullscreen apps.
• Polled WhatsApp Web's DOM via a preload script every 2.5s to detect unread badges without Puppeteer or any external Chromium dependency.
• Shipped as a single portable .exe with a tray icon for runtime contact switching and a config.json for dot size, color, position, and polling interval.