Episode 1 : How to turn your Use Case into an AI pipeline

​

​

​

Episode 2 : How to create a labelled dataset

​

​

​

Episode 3 : How to train an highlight detector

​

​

​

Episode 4 : How to deploy your detector into a lambda function

​

​

​

Episode 5 : How to make your model available to users

​

Episode 1 : How to turn your Use Case into an AI pipeline

• Definition of use case
• Risk assessment
• Select components of an AI pipeline

​

Episode 2 : How to create a labelled dataset

​

​

​

Episode 3 : How to train an highlight detector

​

​

​

Episode 4 : How to deploy your detector into a lambda function

​

​

​

Episode 5 : How to make your model available to users

​

#01 - How to upload custom Pytorch models on Sagemaker

​

​

#02 - How to convert models with ONNX for a lambda deployment

​

​

#03 - How to deploy using AWS batch

​

How to label data

1. Why Labelling data?
1. What is data labelling
2. Why it is important
2. The process
• Define the problem to be solved
• Create a dataset
• Define the labels
• Define the annotation protocol
3. Labelling campaign
• Estimate budget
• Quality control
• Harvesting data
4. ​

Collaterals for videos -

Instructions

Step 1: Write title of the video..

so that the talent knows what is what

Sparkd - Who we are?

EXAMPLE

Place the collateral in this area, so that it is aligned with the eyes of the talent

Step 2: Place the collateral

EXAMPLE

Step 3: Don’t use greens !!

BAD CHOICE!: if you use this, the dots will disappear!!

Step 4: Include headlines and important text from your script..

so that the audience can focus on the most important points

EXAMPLE

Sparkd! ..an AI Business for forward thinking businesses..

EXAMPLE