Portfolio

A collection of my professional and research work in AI, machine learning, geospatial analysis, and infrastructure analytics.

Current Work

Active Development

GeoTasker.ai

AI-powered geospatial storytelling platform. Describe any topic and get a narrated video with maps, animations, data visualizations, and background music. Built with Claude AI, Next.js, FastAPI, and GPU-accelerated rendering.

Claude AIGeospatialVideo GenerationNext.jsFastAPI
Open SourceComing to PyPI

GeoManim

Python library for creating animated geospatial visualizations using Manim. One-line API for animated maps, choropleths, route animations, and more. Install from GitHub: pip install git+https://github.com/Kamol-Roy/geomanim

PythonManimGeospatialAnimationOpen Source

Professional Work

CPCS2021 - Present

Infrastructure Analytics & AI

Geospatial analytics and AI solutions for transportation and infrastructure projects across North America, Saudi Arabia, Central Asia, and Africa. Freight network analysis, travel time reliability, and infrastructure monitoring using large-scale GPS datasets (NPMRDS, INRIX, ATRI).

Built custom Python and Plotly Dash tools for spatial visualization and routing. Led AI-assisted mapping workflows for power and transport infrastructure. Now serving as technical lead for AI and automation initiatives.

PythonPlotly DashGeospatial AnalyticsGPS DataAI/MLInfrastructure

Funded Research Projects

Disruption Identification and Dynamic Mapping During Disasters

2020
NSF Funded

Developed document classification and deep learning models to identify infrastructure disruptions during disasters using streaming Twitter data. Funded by the National Science Foundation (NSF) ORDER-CRISP project.

Deep LearningNLPPythonKerasSklearn
View Project →

Modeling the Dynamics of Hurricane Evacuation

2019
UTC Funded

Built Hidden Markov Models and activity classification systems to understand evacuation behavior patterns using Twitter data, demographic information, and hurricane data. Funded by the Southeast Transportation Center.

Hidden Markov ModelSemi-supervised LearningPythonArcPy
View Project →

Quantifying Human Mobility Resilience

2019
UTC Funded

Developed a novel approach to detect events and measure resilience using geo-tagged Twitter data. Published in EPJ Data Science.

Resilience AnalysisGeospatialPythonEvent Detection
View Project →

Real-Time Traffic Information Sharing via Twitter

2019
FDOT Funded

Analyzed the effectiveness and information content of FDOT Twitter accounts for sharing traffic information. Applied topic modeling and logit models. Funded by Florida DOT.

Topic ModelingLogit ModelPythonRSAS
View Project →

Social Media Crisis Communication During Hurricane Sandy

2020
UTC Funded

Analyzed the efficiency of social media-based crisis communication using entropy analysis, logit models, and regression analysis. Published in the International Journal of Information Management.

Entropy AnalysisRegressionPythonRSAS
View Project →

Hurricane Evacuation Traffic Demand Prediction

2017
UTC Funded

Integrated multiple data sources (geo-tagged Twitter, traffic sensors, socio-demographic data) to predict evacuation traffic using LSTM models.

LSTMMachine LearningMulti-source DataPython
View Project →

Agent-Based Modeling for Human Behavior and Infrastructure Disruption

2020
NSF Funded

Developed agent-based simulations to understand the interaction between human behavior and infrastructure during disasters. Funded by NSF ORDER-CRISP.

Agent-Based ModelingRepast SimphonyJavaSimulation

Risk Perception and Evacuation Decision Modeling

2020
NSF Funded

Built agent-based models to understand how multiple information sources influence risk perception and evacuation decisions during disasters.

Agent-Based ModelingDecision ModelingJavaPython

COVID-19 Projects

Collaboration Projects

Individual & Challenge Projects