Medio da experincia real dos

usurios com sensores e Machine

Learning
by

Vitor

About me
Especialista em Cloud Computing e Integraes,
Vitor um professional que atua tanto no
desenvolvimento de software para plataformas
de alto nvel quanto em sistemas embarcados.
Atualmente atua como Principal Cloud Architect na Gubes e
tambm consultor de Integrao no CrazyTechLabs.

vitormeriat.com.br
linkedin.com/in/vitormeriat
twitter.com/vitormeriat
github.com/vitormeriat

Agenda
The new Things and the IoT
The User Experience
The Machine Learning
The Case
The future

IOT
2010

2016

Medication adherence

Child and elder

monitoring
Alarm
clock

Health monitoring

Sports
and fitness
Refrigerat
Automobil
or
e
COMMUTE

HOME
HOM
E

Pet tracking

Behavior modification

Indoor
Smart
lighting
navigation
VoIP
Computer
Printer
phone

Object tracking

Beacons
Media
and player
proximity

Trip tracking
and car health
Microwave

COMMUTE

WORKPLACE
WORKPLA
CE
Smart appliances

Coffee
maker
Sleep tracking

Home security

Home
automation

Food and nutrition

tracking

Smart
Oven
Air
conditioning
and
scale
temperature control

Leak detection

Office
equipment

Identity

Cell phone

HVAC

Environmental
sensors

Garden, lawn and plant

care

HOME
HOM
E
Smart vending
machines
Vending

Bike ride stats and protection

Security

Information capture

New devices and

sensors

Television
Control

Entertainment systems

O que Experincia de Usurio

Percepo e reaes de uma pessoa
que
resultam do uso ou antecipao de
um produto, sistema ou servio.

Experincia do usurio todo o

processo de relacionamento do
consumidor com o produto

Afeta a Percepo de Valor

Como
resolvemos
nosso
problema de
UX?

Machine Learning
Aprendendo com os dados

WHAT SOCIETY THINKS I

DO

WHAT OTHER
PROGRAMMERS
THINK I DO

WHAT I THINK
I DO

WHAT MY FRIENDS THINK I

WHAT I REALLY DO

from pyspark import SparkContext

from pyspark.mllib.classification import LogisticRegressionWithSGD
from pyspark.mllib.regression import LabeledPoint
#Load and parse data

Intellige
nt
action
What happened?
Why did it
happen?

What will
happen?

Wha

Power BI
Machine Learning
SQL Data Warehouse
HDInsight
Information
management

Big Data
stores

Machine
Learning

Data Lake Analytics

Analytics

Data Lake Store

Stream Analytics
Data Factory
Data Catalog
Event Hubs

Fluxo de dados
1

Coleta e envio dos dados dos

Sensores
4

Device notified (even when app is

PNS
inactive)
(APNS, WNS,
GCM)

Store handle in app backend

Send notification to
handle

How Push Notifications Work

1

Retrieve device
handle
4

Device notified (even when app is

inactive)

Store handle in app backend

PNS
(APNS, WNS,
GCM)

Send notification to
handle

How Service Bus Notification Hub

Works
1

Retrieve device
handle
PNS

Push notification

Registration with tags

Send notification to
handle

Economia compartilhada!
Stopping
instance, stops
billing

$$$$
$$

Per-minute billing
No need to delete
your virtual
machine
Start back with
same disks
attached & same
end-points
22

 Provision Role Instances

Deploy App Code
Configure Network

Network Load Balancer

Windows Azure Datacenter

Network loadbalancer configured

for traffic

Things

Now Available to Purchase

Azure IoT Suite

Device Connectivity and Security
Data Ingestion and Command & Control
Stream Processing & Predictive Analytics
Workflow Automation
Dashboards and Visualization
Preconfigured Solutions

Loosely Coupled

Tracking

Tracking

Store Front
End

Order Queue

Shipping
Service

ASP.NET MVC
4 Mobile App

Service Bus
Queue

Internet
Hi!

 Provision Role Instances

Deploy App Code
Configure Network

Network Load Balancer

Windows Azure Datacenter

Network loadbalancer configured

for traffic

Partner IoT capabilities

Systems Integration

Devices

Devices

Vertical applications

Systems integration

Analytics
(Data Scientists)

Analytics
(Data Scientists)

Vertical applications

Azure IoT Reference Architecture

Devices and
Data
Sources

Data
Transport

Device and Event Processing

Provisioning API

IP capable
devices
Agent
Libs
Existing IoT
devices
Agent
Libs
Low power
devices

Presentation

Solution Portal

Identity & Registry Stores

Device State Store
Cloud
Gateway
Gateway
Agent
Libs

Data
Visualization &
Presentation

Stream Event Processor

Storag
e

Analytics/
Machine
Learning

Control System Worker Role

Pattern: Telemetry first

Start with
telemetry.
The important
data may not
be what you
expected.
Address
privacy,
security and
manageability
before moving

It is very hard to predict in advance what

data will be useful.

It is tempting, but likely inefficient to try
for business transformation in the first
step.
Think about not only device telemetry
but also diagnostic telemetry.
Privacy and security implications of
telemetry are generally lesser than for
command and control.

IoT architecture requirements

Handle extreme hardware and software heterogeneity.
Lower barriers to entry: evaluate -> prototype -> deploy.
Provide hot-path and cold-path analysis and response.
Build for hyper-scale and enable low latency.
Be secure by design; support defense in depth.

IoT solutions until now

Most of the early successful
IoT deployments were
either
For very complex and
expensive devices, where the
cost of a custom
hardware/software solution is
acceptable compared to the
cost of the device, or
For high-volume,
homogeneous devices, where
the software needs are

Device complexity

<h1> Obrigado! </h1>

</body>
</html> <!-- The End -->
vitormeriat.com.br
linkedin.com/in/vitormeriat
twitter.com/vitormeriat
github.com/vitormeriat