pascal-holy
commented
3 years ago Proposed solution:
- Minimal Terraform setup that launches airy core via Helm in the users AWS account
- Kafka image that contains preloaded data
- Metabase pod launched and configured automatically to access the same data from a preloaded data lake bucket
Limitations:
- Users need an AWS account with enough permissions to create EKS cluster, S3 Bucket and several IAM roles
- There is no live connection between Kafka/Inbox and the data-lake
Stretch Goals
- Jupyter Notebook connected to the data lake to play around with the data
As a data scientist or engineer I want to experience the power of Airy's conversational analytics dashboard (powered by Metabase) in a quick 20min demo.
This will be the diy analytics project that shows off the power of the Airy platform and our conversational analytics know how, described in https://blog.airy.co/introducing-data-lakes-for-conversational-data/ https://blog.airy.co/a-guide-to-conversational-metrics/ https://blog.airy.co/how-to-build-your-conversational-dashboard/
The demo repo should include everything a data scientist needs to:
Set up their Airy Instance
populated with (real-ish looking, so perhaps GPT-2 generated) conversations in it
have a running Metabase dashboard running in the end with the following metrics: -- Number of Conversations -- Active Conversations by Day and Source -- Most active Hours -- Channels -- Average Human Response Time -- Conversation Count -- Conversations responded by Bot -- Conversations responded by Humans
Deliverables
A good comparison: Good example Machine learning Repo that comes with a data set, a 20min video, in-depth explanations and guide https://github.com/kaiwaehner/hivemq-mqtt-tensorflow-kafka-realtime-iot-machine-learning-training-inference