1 of 8

GPU-Accelerated GeoAI (HeavyIQ)

for Geospatial Big Data Processing

Devika Jain (Harvard CGA)

Dr. Michael Flaxman (HeavyAI)

2 of 8

SQL

Fully integrated, fast and powerful database querying

VIZ

Fully integrated, fast and powerful database querying

GEO

Geospatial Intelligence expanding data discovery

ML

Empowering smarter ecosystems with machine learning.

AI

Conversational intelligence driven by natural language

What is

HEAVY.AI?

3 of 8

Technology Stack & Integrations

Platform Architecture

4 of 8

Conversational Visual Analytics

HeavyIQ Workflow

Ask Question

Ask a question about your data

t

1

Translate

Text to SQL

t

2

Data Retrieval

Execute SQL with HeavyDB

t

3

Results

Multiple ways to express insights

t

4

Data Visualization

Natural Language

Tabular

5 of 8

HeavyIQ: Interactive Conversational Analytics at Scale

6 of 8

  • Global record of tweets spanning time, geography, and language; extending from 2010 to the present; about 10.3 Billion�
  • A comprehensive collection of geolocated tweets available to the research community, stored on Harvard’s High Performance Computing Cluster �
  • Provides easy access to both the dataset and processing tools in an integrated environment for Harvard users�
  • Useful not only for geospatial research but also research involving Social Science, Business and Humanitarian studies

CGA’s Geotweets Archive

7 of 8

8 of 8

HeavyAI / IQ

Next Steps

Available Today @ CGA (large archive)

https://gis.harvard.edu/geotweet-archive-v20

https://gis.harvard.edu/geotweet-request-form

�Live Online Demo @ Heavy.AI �(smaller archive)

​

“Free Edition” �Downloadable No Code UX

​

“Open Source”

Core GPU database

​

​

​

Image credit: Wired Magazine