✨ stellaephile - Sonal Ganvir

Data that knows
where it stands.

I'm a GIS Data Analyst specializing in urban transport and mobility, currently at WRI India. I work at the intersection of spatial data, AI, and city systems - building tools that help decision-makers see what maps alone can't tell.

My background spans a BTech in Computer Science from IIIT Nagpur and a Master's in Data Analytics from the University of Houston, with hands-on experience in GIS platforms, cloud infrastructure, and generative AI applications.

I build on GCP and AWS, compete in Gen AI hackathons, and take on freelance GIS projects that push the boundaries of spatial analysis.

I believe a city is a living thing - and every dataset it generates is part of its breathing.

✦ Skills

QGIS ArcGIS OpenStreetMap GTFS PostGIS Python SQL Pandas GeoPandas Tableau Google Cloud AWS Bedrock Vertex AI LangChain Gen AI

Years in Data Analytics & GIS

Projects Across GIS, ML & Gen AI

Gen AI Hackathons Competed

∞

Cities Still Left to Map

Projects that navigate
complexity.

GCP · GTFS · ADK AI Agent

Propus AI - Transit Intelligence Agent

An AI agent connecting Delhi GTFS and OpenStreetMap data to answer real-time transit queries. Built on Cloud Run for the Google APAC Academy.

↗ Demo ↗ GitHub

Streamlit · Vertex AI · GCP · Cloud Run

Wayfarer AI - Smart Travel Booking

Built at the Google GenAI Exchange Hackathon 2025. A conversational AI travel planner that personalises itineraries, compares transport modes, and gives real-time budget-aware recommendations for sustainable travel.

↗ GitHub

Gen AI · NLP · PostGIS

What's Up Doc

An intelligent healthcare navigation app that performs a brief symptom severity assessment and surfaces nearby medical facilities tailored to your needs.

↗ GitHub

Deep Learning · CNN · ANN · IEEE

Automatic Speaker Recognition - Transfer Learning

B.Tech capstone published in the IEEE Journal. Pioneers ASR using transfer learning on Librivox audiobooks - 200 speakers, 19K spectrogram images for CNN, 28K spectral samples for ANN.

↗ GitHub

Audio ML · VGGish · Random Forest

IndiSpeak - Multilingual Audio Classification

Classifies audio embeddings into diverse Indian languages using VGGish model extraction and Random Forest classification, with GridSearchCV hyperparameter tuning for precision.

↗ GitHub ↗ Slides

Deep Learning · Azure · Web App

Bhaasha - Indic Language Detection from Text

Deep learning web app detecting 10 major Indian languages in text with 80%+ accuracy. Trained on MultiIndicMT dataset, with Azure Hyperdrive tuning and ISCII character encoding.

↗ GitHub

Data Science · Sports Analytics

Cricket Unleashed: Decoding the Game with Data Science

Medium · @sonalgan2000

Mapping the unseen
patterns of the world.

Data that knows
where it stands.

Projects that navigate
complexity.

Propus AI - Transit Intelligence Agent

Wayfarer AI - Smart Travel Booking

What's Up Doc

Automatic Speaker Recognition - Transfer Learning

IndiSpeak - Multilingual Audio Classification

Bhaasha - Indic Language Detection from Text

Thinking out loud
on Medium.

Let's map something
worth building.

Mapping the unseenpatterns of the world.

Data that knowswhere it stands.

Projects that navigatecomplexity.

Propus AI - Transit Intelligence Agent

Wayfarer AI - Smart Travel Booking

What's Up Doc

Automatic Speaker Recognition - Transfer Learning

IndiSpeak - Multilingual Audio Classification

Bhaasha - Indic Language Detection from Text

Thinking out loudon Medium.

Let's map somethingworth building.

Mapping the unseen
patterns of the world.

Data that knows
where it stands.

Projects that navigate
complexity.

Thinking out loud
on Medium.

Let's map something
worth building.