CTO Straight Talk - Issue 2 - 37

We need to make machine learning more accessible - to
every enterprise and, over time, everyone.

but also in enterprises, changing the

subsidiaries. I came to Microsoft in

of predictive analytics dramatically

way they work. For example, Thyssen

2013 because the CEO gave me the

faster and cheaper, and is leading

Krupp Elevators, a customer of ours,

opportunity to realize a new vision

them to trustworthy and actionable

wants to know ahead of time when its

for making machine learning a

business insights on big data.

elevators are ready for maintenance

mainstream application that can be

in order to deploy maintenance

used by all, leveraging the cloud.

crews more effectively and minimize

For example, eSmart Systems of
Norway is pioneering smart grid

We have a vision of creating the

management using our tools. A

Intelligent Cloud, which would make

traditional smart grid - an electricity

machine learning more accessible

supply network that uses digital

to every enterprise and, over time,

communications to detect and react

every one of us. Machine learning

to local changes in usage - includes

When combined with machine

software today is usually managed on

multiple data silos, including SCADA

learning, it turns out this kind of

premises by each organization using

networks, building automation

data can help you spot malfunctions

it, and building machine learning

systems, and substation meters. In

that are going to happen. Machine

applications requires expert data

this environment, it can be difficult

data analyzed by machine learning

scientists. However, data scientists

to forecast consumption and prevent

algorithms improves system

are in short supply, commercial

bottlenecks or outages. For a utility

reliability by determining the

software licenses can be expensive,

company, upgrading its entire

probability of failure well ahead of

and popular programming languages

infrastructure would be costly. Even

time and, eventually, making sure

for statistical computing have a steep

when upgrades are made, with new

most things that matter never fail.

learning curve. Even if a business

smart sensors or meters, data often

could overcome these hurdles,

gets collected but is not readily

deploying new machine learning

accessible. eSmart Systems is now

models in production often requires

using our cloud platform to integrate

months of engineering investment.

and analyze usage data and create

Scaling, managing, and monitoring

forecasts. Azure Machine Learning is

these production systems require the

the "brains" of their solution, running

capabilities of a very sophisticated

the data models for predictive

engineering organization, which few

analytics. The analytics are used to

enterprises have today.

predict capacity problems and

elevator downtime. It now relies
on predictive analytics for this
information, based on data provided
by sensors in the elevators.

Realizing a New Vision for
Machine Learning
Since I was 12 years old, I have
wanted to work with artificial
intelligence. I did a PhD in neural
networks, and my first job involved
creating a machine learning
application for credit card fraud

automatically control load in

detection. At Amazon, I spent nine

Microsoft's Azure Machine Learning

years in a variety of roles, but I most

is helping organizations meet those

enjoyed building and running the

challenges. Combined with the rest

Mendeley is another innovative

company's central machine learning

of Microsoft's data platform, it allows

customer. One of the biggest

group and being responsible for risk

our customers to create entirely new

repositories of scientific research

management for Amazon and its

solutions that deliver on the promise

content in the world, Mendeley

individual buildings.

CTO Straight Talk | 37
CTO Straight Talk - Issue 2

Table of Contents for the Digital Edition of CTO Straight Talk - Issue 2
