Akasi | Data Concierge


The Challenge | Data Concierge

Develop an artificial intelligence tool to help Earth science data users and enthusiasts find datasets and resources of interest!


No Video Provided

Akasi aims to get the right data to users of all walks of life in a convenient way.


The team were originally drawn to this project because it involved machine learning, user-centered design, and because it matched well with IBM Watson services which the team were eager to learn and implement.

For the team the main parts of the challenge were:

  • Creating an easy to use interface for people from all walks of life
  • Applying machine learning to improve the system for users

The team were keen to try building a chat bot because it has the advantageous position of using channels users are familiar with, and is a more personal interactive experience than other formats for accessing NASA data. The name Akasi was chosen, it is a female Indian name meaning "atmosphere". It conveyed the personable environmental focus we wanted our App to convey.

A key challenge was how to implement botmaster. In the UK an open source framework called botmaster is advised as the best way to integrate a chat bot with different channels and external API calls. However, many hours were wasted as the documentation was incorrect for a facebook integration. The project moved to twitter, however the going was very slow. The team also needed to re-assess the main purpose of the application and decided that the emphasis shouldn't be on as much data as possible, but on making NASA's data easy to access.

The team went forward with a new focus and the decision to build more quickly using NodeRED.

Building in nodeRED has led us to make use of

  • Cloudant DB
  • Watson Conversation
  • Twitter API
  • Google Geocode API
  • Data & Analytics Insights for Twitter
  • Data & Analytics Weather Company Data
  • Watson Personality Insights
  • CloudFoundry NodeJS SDK

Coupled with the Bluemix chain is a nodeJS server that provide an API to interface between the Watson system and the NASA GIBS Tile server for MODIS visual and Aura NOX data. The relevant MODIS imagery is derived from the geo-information from the Twitter interactions whilst the NOX and other air quality data sets are thresholded to provide a traffic light overlay to the user. The Jimp graphics library was used in the nodeJS to provide image processing functionality.


SpaceApps is a NASA incubator innovation program.