Developing Sophisticated Serverless Applications with AI

© 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Adrian Hornsby, Technical Evangelist
http://inserverless.com - 2017
Developing Sophisticated
Serverless Applications with AI

• Technical Evangelist, Developer Advocate,
… Software Engineer
• Own bed in Finland
• Previously:
• Solutions Architect @AWS
• Lead Cloud Architect @Dreambroker
• Director of Engineering, Software Engineer, DevOps, Manager, ... @Hdm
• Researcher @Nokia Research Center
• and a bunch of other stuff.
• Climber, like Ginger shots.

What to expect
• Quick intro
• 3 demo applications
• Polly
• Rekognition
• MXnet
• Wrap up.

AI and Serverless
…like Salt and Pepper

Event driven
A B CEvent on B by A triggers C
Invocation
Lambda functions
Action

How Lambda works
S3 event
notifications
DynamoDB
Streams
Kinesis
events
Cognito
events
SNS
events
Custom
events
CloudTrail
events
LambdaDynamoDB
Kinesis S3
Any custom
Invoked in response to events
- Changes in data
- Changes in state
Redshift
SNS
Access any service,
including your own
Such as…
Lambda functions
CloudWatch
events

Event-driven using Lambda
AWS Lambda:
Resize Images
Users upload photos
S3:
Source Bucket
S3:
Destination Bucket
Triggered on
PUTs

No servers to provision
or manage
Scales with usage
Never pay for idle Availability and fault
tolerance built in
Serverless means…

EVENT DRIVEN CONTINUOUS SCALING PAY BY USAGE
Serverless means…

The rise of AI

The Advent of AI & Deep Learning
Data
GPUs
& Acceleration
Cloud
Computing
Algorithms
AWS

AI On AWS Today

AI & Deep Learning in the hands of every developer

© 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.© 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Voice enabled applications

Polly: Text In, Life-like Speech Out
Amazon Polly
“Today in Seattle, WA
it’s 11°F”
“Today in Seattle Washington
it’s 11 degrees Fahrenheit”

https://www.w3.org/TR/speech-synthesis/
<speak>
The spelling of my name is
<prosody rate='x-slow'>
<say-as interpret-as="characters">Adrian</say-as>
</prosody>
</speak>
A Focus On Voice Quality & Pronunciation

<API>
Amazon Polly
</API>
aws polly synthesize-speech
--text "It was nice to live such a wonderful live show"
--output-format mp3
--voice-id Joanna
--text-type text johanna.mp3

Developing Sophisticated Serverless Applications with AI

<demo>
PollyCast
</demo>
* Initial project by James Siri, Piotr Lewalski
https://github.com/adhorn/pollycast

Image analysis

Amazon
Rekognition
Object and scene detection
Facial analysis
Face comparison
Celebrity recognition
Image moderation
Rekognition: Images In, Rich Metadata Out

<API>
Amazon Rekognition
</API>
aws rekognition detect-faces
--image '{"S3Object":{"Bucket":"adhorn-reko","Name":"horse.jpg"}}'
--attributes "ALL"
aws rekognition detect-labels
--image '{"S3Object":{"Bucket":"adhorn-reko","Name":"horse.jpg"}}'

Detect labels with
Rekognition

Add
Rekognition tags

Extract Image
Metadata

AWS Step Functions:
Orchestrate a Serverless processing
workflow using AWS Lambda

<demo>
Image Recognition and Processing Backend
Step Functions
</demo>
https://github.com/awslabs/lambda-refarch-imagerecognition

<demo>
Poliko
powered by Amazon Polly & Rekognition
</demo>
http://poliko.adhorn.me

Demo - Poliko
Poliko
Take Pic
Amazon Cognito
2. Detect Labels
4. Synthesize-speech
Amazon Rekognition
Amazon Polly
3. Detect Faces
Amazon S3
“Static website hosting” enabled

Deep Learning enabled applications

One-Click
Deep Learning
AWS Deep Learning AMIs
Amazon Linux & Ubuntu
Up to~40k CUDA cores
Apache MXNet
TensorFlow
Theano
Keras
Caffe
CNTK
Torch
Pre-configured CUDA drivers
Anaconda, Python3
Out-of-the-box Tutorials
+ CloudFormation template
+ Container Image
Available in the AWS Marketplace
AI Frameworks on AWS

0.2
-0.1
...
0.7
Input Output
1 1 1
1 0 1
0 0 0
3
mx.sym.Pooling(data, pool_type="max", kernel=(2,2), stride=(2,2)
lstm.lstm_unroll(num_lstm_layer, seq_len, len, num_hidden, num_embed)
4 2
2 0
4=Max
1
3
...
4
0.2
-0.1
...
0.7
mx.sym.FullyConnected(data, num_hidden=128)
2
mx.symbol.Embedding(data, input_dim, output_dim = k)
Queen
4 2
2 0
2=Avg
Input Weights
cos(w, queen) = cos(w, king) - cos(w, man) + cos(w, woman)
mx.sym.Activation(data, act_type="xxxx")
"relu"
"tanh"
"sigmoid"
"softrelu"
Neural Art
Face Search
Image Segmentation
Image Caption
“People Riding Bikes”
Bicycle, People,
Road, Sport
Image Labels
Image
Video
Speech
Text
“People Riding Bikes”
Machine Translation
“Οι άνθρωποι
ιππασίας ποδήλατα”
Events
mx.model.FeedForward model.fit
mx.sym.SoftmaxOutput
Anatomy of a Deep Learning Model
mx.sym.Convolution(data, kernel=(5,5), num_filter=20)

Early detection of diabetic
complications

Autonomous Driving Systems

Real Time, Per Pixel Object Segmentation

Pre-trained MXNet
models
http://data.mxnet.io/models/imagenet/

ResNet – Deep Learning
Based on:
Deep Residual Learning for Image Recognition
Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
https://arxiv.org/pdf/1512.03385.pdf
https://medium.com/towards-data-science/neural-network-architectures-156e5bad51ba

Scale Predictions with AWS Lambda and MXNet
AWS LambdaAmazon
API Gateway
Amazon S3
Training
Inference
https://aws.amazon.com/blogs/compute/seamlessly-scale-predictions-with-aws-lambda-and-mxnet/

Predict with AWS Lambda

AI for everyone!

Thanks you!

Developing Sophisticated Serverless Applications with AI

More Related Content

Similar to Developing Sophisticated Serverless Applications with AI (20)

More from Adrian Hornsby (18)

Recently uploaded (20)

Developing Sophisticated Serverless Applications with AI