►
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
B
Hello,
hello,
can
you
hear
me
good,
hey,
hi,
everyone,
I'm,
the
shark,
a
turkey,
a
product
manager
and
OpenShift,
a
IML
on
OpenShift
I
focus
on
that
I'm
very
pleased
to
introduce
you
this
next
topic,
which
is
about
really.
How
do
you
do
machine
learning
automated
and
on
top
of
OpenShift
with
us
today
we
have
a
tie
and
we
have
guy
from
the
IDF
and
they're
going
to
start
next.
Thank.
A
You
sure
so
hello,
everybody
and
we're
Italian
guy
from
the
IDF
and
today
we're
gonna
talk
about
machine
learning
platform
we
developed
on
top
of
openshift
and
kubernetes
that
create
state-of-the-art
machine
learning
models
and
utilize.
The
data
scientists
and
software
engineer
jobs
in
our
organization.
So
a
little
bit
about
us
I
meet,
I
I'm
the
machine
learning
team
leader
in
the
IDF
and
actually
everything
you
see
here
in
the
demo
is
something
that
we
built
in
our
team
and
I
I'm.
C
Guy
I'm
I'm,
the
manager
of
a
private
cloud
managed
services
project,
especially
openshift
in
that
case
that
we're
gonna
see
today
so
a
little
bit
about
IDF,
the
Israeli
defense,
all
wear
a
tie
and
I
come
from
right
now.
This
very
Defense
Force
are
in
the
process
of
digital
transformation
that
our
main
goal
is
to
accelerate
the
delivery
and
the
development
of
application
inside
our
systems.
C
A
So
an
atom
of
removal
to
the
main
topic,
how
we
can
make
you
each
already,
each
and
everybody
here
in
this
room,
machine
learning,
expert
and
build
a
really
state-of-the-art
machine
learning
models
in
just
minutes
or
hours.
But
first
we
need
to
understand
a
little
bit
about
machine
learning
and
its
basics.
So
actually
machine
learning
is
just
learning
from
previous
data
in
order
to
predict
the
future
and
we'll
follow
up
in
examples
that
will
demonstrate
the
word
process
during
all
the
presentations.
A
So
we
have
you
a
data
set
of
medical
diagnoses
and
there
was
our
mission-
is
to
predict
if
James
will
have
flu
or
not
based
on
these
parameters
that
we
can
see
here
and
how
we
can
do
it.
Actually,
we
have
a
wide
variety
of
machinery
models
and
algorithms.
This
is
a
release,
a
small
sample
of
them
and
we
can
use
each
and
every
one
of
them
in
order
to
predict
and
create
models
that
help
us
to
predict
the
future.
A
But
each
model
here
has
its
own
configuration
or
on
his
parameters
and
it's
usually
a
very
exhausting
task
to
just
select
the
best
one
and
fit
it
to
the
data.
So
we
need
help
because
it's
actually
can
take
a
lot
of
math
a
lot
of
time
in
order
to
find
the
best
solution
to
the
problem,
and
so,
let's
deep-dive
about
what
the
data
scientists
really
do
when
he
gets
a
new
data
set
and
start
building
the
models.
So
first
we
start
with
the
data.
A
We
have
a
step
called
the
data
engineering
which
include
removing
irrelevant
columns,
making
the
data
more
predictable.
Like
you
know,
flu
example
will
obviously
remove
the
patient
name
column,
because
it's
obviously
will
not
predict.
If
one
will
get
a
flu
or
not,
but
we
will
we
will.
We
assume
that
fever,
for
example,
will
give
us
more
predictive
analysis
if,
if
about,
if
one
will
get
or
not
so
we'll,
obviously
we'll
give
it
more
weight.
A
For
example,
after
that
we
start
the
machine
learning
task,
which
is
just
taking
a
lot
of
algorithms
and
models
and
start
fitting
them
to
the
data
set
to
the
problem.
In
order
to
get
the
predictive
model,
you
can
see
that
it's
a
cycle
and
it's
a
almost
usually
it's
almost
everytime,
exhausting
process
that
takes
a
lot
of
time.
A
It's
based
on
trial
and
error,
so
it
can
take
even
a
few
months
for
specific
data
set,
and
after
that
we
have
the
operations.
We
need
to
serve
this
model
to
as
a
production
service
to
different
applications
and
consumer
to
consume
the
release.
The
result
of
this
model
to
predict
the
future.
So
an
overestimation
of
you
know
in
an
organization,
for
example,
if
we
take
100
data
sets,
new
data
sets
that
get
into
our
organization
in
the
final
round.
A
Only
five
are
really
going
to
production,
and
that's
because
of
this
exhausting
process
of
building
machinery
models.
Some
of
them
are
falling
because
of
really
really
small
things
that
and
it's
just
a
waste
of
data
and
waste
of
knowledge.
Really
so
from
this
pipeline,
we
can
gather
four
different
top
challenges
that
we
are
coming
to
solve.
The
first
one
is
environment.
A
We
saw
that
we
have
a
cycle
of
Fame
machine
learning,
building
process
and
each
time
this
cycle
includes
a
lot
of
model
evaluations
that
are
usually
not
stored
everywhere.
So
we're
no
key
they're,
not
keeping
track
of
results.
You
just
got
for
each
model
and
it
comes
figuration
of
model,
and
this
is
not
good,
because
it's
not
only
to
waste
a
lot
of
time.
We
can
use
this
history
later
in
other
projects
and
experiments
in
order
to
make
the
building
process
more
efficient
and
we're
gonna
see
it
also
in
the
demo.
A
The
third
one
is
optimization
and
we
just
want.
We
look
for
a
tool
that
just
take
a
wide
wide
search
space
space
and
just
give
me
the
best
combination
of
search
results
of
just
to
give
me
the
best
mode,
obviously,
and
we
said
how
we
solve
it,
distributed
on
openshift
and
how
we
utilized
OpenShift
into
it
efficiently
and
the
last
one
is
deployment,
and
here
we
have
a
gap
between
data
scientist,
scientist,
knowledge
and
software
engineers,
knowledge,
because
the
scientist
doesn't
know
enough
about
docker
writing
applications
and
the
operations.
A
So
we
can't
just
take
his
mathematical
model
and
deploy
to
production
as
the
rest
api
and,
on
the
other
hand,
the
software
engineer
doesn't
know
how
to
handle
this
mathematical
model
that
I
said
that
the
scientist
built
and
expose
it
as
a
rest
api.
So
this
is
why
a
lot
of
models
really
just
don't
go
up
to
production
and
we're
gonna
see
how
we
solved
it.
So
now,
after
we
understood
the
challenges,
let's
see
how
we
solve
them.
So
basically,
let's
go
back
to
the
first
challenge.
A
It
was
environment
and
we
deployed
jupiter
hub,
which
is
a
common
tool
of
data
scientists
these
days
and
we
control
our
this
resources.
Our
resources
are
being
allocated
it's
allocated
and
dynamically.
Actually
so
guys,
let's
just
spawn
a
new
notebook,
and
here
you
can
see
a
different
variety
of
of
machine
learning.
Let's
say
environments.
A
Actually,
each
notebook
is
just
a
docker
image
that
is
being
controlled
by
us
and
when
the
research
is
over,
when
the
data
signs
go
to
sleep
and
on
that
day
he
just
turn
off
the
computer
turn
of
the
notebook
and
the
resources
has
been
freed
to
other
data
scientist.
This
includes
also
GPUs,
which
are
we
have
everybody
in,
doesn't
have
enough
GPUs
today.
A
So
this
is
how
we
resolve
this
thing,
and
now,
let's
move
on
to
my
environment
that
I
prepared
before-
and
you
can
see
here,
Jupiter
with
the
flu
data
set,
we
have
just
generated.
We
have
here
the
same
data
that
we
saw
in
the
example
and
a
notebook
that
includes
our
demonstration.
So
what
I'm
going
to
do
now?
We're
just
trying
to
fit
the
data
to
machine
learning
models
so
we're
going
to
run
a
different.
The
basic
data
science
type
data
science
operations
in
order
to
fit
the
data
Tuesday
to
it.
A
So
now
we're
just
reading
the
data
we
are
dropping
irrelevant
columns.
Like
the
name
column,
we
are
just
converting
some
columns
to
numbers
and
dates,
so
the
mathematical
algorithm
can
understand.
What's
in
the
data
and
splitting
it
in
order
for
us
to
evaluate
it
now,
let's
say:
remember
the
second
problem:
the
history
problem
I'm
going
to
build
you
now,
three
decision,
trees,
algorithms,
a
decision
tree
is
just
a
predictive
model
that
lends
the
data
ends
and
know
how
to
predict
future
data.
A
So
it
has,
we
can
control
the
depth
of
the
tree
and
this
parameter
actually
really
influence
the
performance
of
the
model.
So
thank
one
to
create
three
different
decision
tree
classifier
and
we're
going
to
change
this.
This
much
depth
parameter
every
time.
So
we
start
with
a
depth
of
depth
depth
of
three
and
we
got
0.54
accuracy.
A
A
So
now
who
remembers
the
value
of
depth
equal
to
three
I
assure
that
we,
with
the
concert,
a
concentrated
enough,
is
remembering
it
now,
but
if
I
run
another
1000
experiments
I
assure
you
that
no
one
will
remember
what
was
the
result,
and
maybe
it
was
a
good
result
and
we
are
losing
a
lot
of
data
that
way
so
right
now,
I'm
going
to
show
you
how
we
solved
it
using
our
ever
our
ml
tracker
platform,
which
is
also
hosted
on
openshift.
So
let's
just
go.
A
This
is
the
UI
of
this,
and
then
we
are
going
to
create
a
new
project.
You
can
see.
I
just
need
to
specify
the
project
name
and
a
description
and
I
can
control,
also
using
node
select
all
feature
of
kubernetes
which
resources
actually,
which
which
resources
will
be
running
the
pods
and
the
workloads
that
my
experiment
needs.
A
A
You
see
an
empty
project
right
now
and
let's
return
to
the
Jupiter
and
type
in
a
newly
created
project
here
and
we
are
doing
the
same
thing-
we're
just
running
a
decision
tree
algorithms
with
different
parameters.
You
can
see
that
I'm
importing
here
a
Python
package,
we
wrote
and
we
will
use
it
during
the
wall
representation,
so
we're
just
creating
a
track
here
and
we
know
how
to
when
we
input
it's
the
model
object
we
give
as
input
also
the
accuracy
skull
and
each
key
value
matrix.
A
And
move
back
to
the
tracker
and
we're
gonna
see
that
we
have
three
different
experiments
with
the
results
and
the
matrix.
We
can
see
that
we
know
how
to
extract
each
and
every
parameters
of
the
model,
and
it
helps
us
in
the
future
to
understand
which
model
was
really
the
best
best
one.
And
so
this
is
the
history
part.
A
That's
actually
going
to
run
my
models
and
here
and
if
I'm,
defining
a
safe
space
right
now,
I'm
searching
between
two
different
algorithms
and
decision,
trees
and
cannons,
and
each
algorithm
has
its
own
parameters
right
now.
We're
just
testing
two
different
parameters:
each
algorithm-
and
you
can
see
the
range
here.
A
We
actually
define
defining
a
really
large
search
space
for
our
optimization
problem,
for
example
the
much
depth
what
we
tested
before
right
now,
we're
testing
it
between
with
the
range
of
3
and
20
and
we're
going
to
define
it
and
then
deploy
the
optimization
task
to
OpenShift
and
this
platform.
A
good
thing
to
mention
is
that
we
are
not
just
running
through
the
all
combinations.
We
have
an
s,
smart
search,
algorithm.
That
knows
how
to
do
it
really
really
fast,
and
you
can
pay
attention
to
what
we
need
to
specify
here.
A
So
it
just
displays
the
objective
function
we
defined,
which
search
algorithm
I'm,
going
to
use
the
number
of
workers
which
will
obviously
see
late
later.
What
what
does
it
mean
and
how
many
evaluations
I
want
to
do?
That
means
how
many
models
I
will
actually
build
and
the
more
the
better,
but
it
will
it
will
it's
also
a
power
consuming.
So
now,
let's
run
it
and
see
behind
the
scene
in
the
openshift.
What's
going
to
happen,.
A
A
And
move
back
to
up
and
shift-
and
you
can
see
here
that
a
new
pod,
the
manager
pod
is
being
created
and
after
it
three
different
machine
learning.
Workers
are
created
because
we
specified
with
before
and
then
the
responsibility
is
to
just
run
models
with
different
combination.
The
dist
manager
is
responsible
to
provide,
and
then
you
can
see
we
just
specify
100
evaluations
and
in
about
five
seconds.
A
So
we
I'm
actually
not
going
to
show
you
how
we
deploy
real
models
but
I'm
going
to
show
you
what
is
needed
in
order
to
to
deploy
one.
So
it's
we
only
need
to
specify
which
frameworks
the
model
was
built
in
the
path
to
where
it
is
day.
Well.
Is
it
saved
as
a
file,
and
we
support
an
object,
storage
today
and
the
pre-processing
function?
Name
that
is
responsible
for
converting
the
data.
We
want
to
predict
to
a-1.
That
model
can
understand
using
the
same
basic
data
science
operations
we
did
earlier
after
we
run
it.
A
C
C
C
The
second
and
main
component
of
this
architecture
is
the
GPU
compute
nodes,
each
compute
node
of
a
2v
100
nvidia
cards
on
it,
especially
for
all
the
GPU
were
closed,
that
the
system
needs
to
be
done
application,
and
actually
we
have
a
quite
problem
there
cause
right
now,
each
pod
can
utilize
can
assign
GPU
and
no
other
pods
can
share
the
GPU
with
it.
So
when
you
assign
to
one
pod
one
GPU,
the
GPU
is
only
for
that
pod
and
actually
on
workload
in
production.
You
don't
utilize
100%
of
the
GPU
and
it's
quite
problem.
C
This
is
the
Nvidia
device
plug-in
that
using
right
now
we
folk
that
plug-in
and
write
our
own
written
plugin
for
the
GPU
scheduled
actually
split.
The
GPU
and
time
share
with
everyone
with
the
other
pods
on
the
system,
and
this
is
our
main
component
of
the
GPU
right
now
we
are
adding
4
4
multi-gpu
node,
based
on
NVIDIA
GTX.
Actually,
when
you
want
to
run
more
complex
machine
land
models,
you
need
to
run
it
on
a
multiple
GPUs,
more
than
one
or
two,
maybe
eight
six
at
a
time.
C
A
So
auto
ml,
and
until
now
we
just
saw
optimization
smart
development
of
how
we
optimize
the
data
scientist
life
in
our
organization,
but
actually
we
needed
some
knowledge
in
coding
and
some
knowledge
in
data
science
in
order
to
use
it.
But
right
now,
I'm
going
to
show
you
how
we
build
machine
learning
models
using
only
a
data
set
and
making
all
this
process
at
automatic.
A
You
can
see
that
we
know
how
to
extract
the
relevant
columns
from
the
fat
from
the
file.
We
just
need
to
specify
again
a
project
name
and
when
we
click
the
interested
column.
In
our
case,
it's
the
got
blue
column.
What
is
happening
behind
the
scenes?
We
are
trying
a
lot
of
different
machine
learning
processes
really
complex,
one
that
scares
the
data
and
make
it
more
predictable.
A
And
after
that
again
we
are
using
the
same
optimization
tool
we
used
earlier,
but
right
now
we're
running,
really
complex
models
that
usually
requires
a
lot
of
optimization
and
consumer
out
of
time.
In
order
to
do
so,
if
you
move
back
to
the
OpenShift,
we
can
see
that
right
now
we
have
only
two
workers
that
Rand
is
this
task,
but
each
one
of
them
is
really
strong.
A
Actually,
it
has
eight
cores
and
16
gigs
of
RAM
in
order
to
do
this
task,
because
it's
inside
this
there
are
really
complex
models
that
built
in
it
may
take
a
while.
It
may
take
an
hour
or
even
more.
It
depends
on
the
size
of
the
data
and
it's
complex,
complex
stability,
but
we
already
prepare
the
deployment
that
we
ran
earlier
and
we're
gonna
show
just
a
REST
API
that
is
going
to
predict
us
if
one
will
get
a
flu
or
not
based
on
the
only
relevant
columns.
A
And
right
now
we
are
going
to
just
give
it
input,
different
values
and
see
the
predicted
result,
and
so,
let's,
let's
type
in
summary,
some
values.
Actually
this
is
the
UI
for
testing
purposes
only
and
real
applications.
That
was
the
other
applications
that
hosted
on
open
shift
also
just
requested
this
prediction
as
using
REST
API.
So
this
is
how
we
we
really
make
smart
applications.
A
Okay,
let's
click
on
to
addict
and
see
this
person
will
get
flow.
We
think,
but
let's
see
how
it,
how
is
it
shown
in
our
platform?
So
let's
move
on
to
the
tracker
again,
we
still
at
0%,
but
it
can
take
a
while.
It
may
take
a
while
and
move
to
our
preferred
project,
and
you
can
see
here
that
we
got
a
new
deployment
with
green
status
and
we
can
also
choose-
and
it's
a
URL,
because
we
support
and
that's
a
kind
of
a
be
testing
for
models
and
we
support
for
each
project.
A
A
It's
a
point
to
mention
and
let's
see,
what's
really
behind
the
scenes
and
what
we
have
in
our
open
chip
project.
So
actually
we
have
scalable
deployments
of
models,
we
have
the
Jupiter
hub
and
we
have
a
lot
of
notebooks
that
being
spawned
on
Jupiter.
Now
we
have
mini,
or
as
that
is
an
object,
storage
that
helped
us
to
save
models
and
crypt
racks
of
expert
experiments.
We
have
a
PostgreSQL
that
helps
to
host
to
be.
A
This
is
a
database
of
our
application
actually
and
every
time
you
cast
MQ
cluster,
and
that
is
actually
the
master
and
the
workers
from
the
optimization
task
actually
communicate
through
a
cluster
rabbitmq
cluster,
a
queue
in
order
to
for
us
to
keep
stable
communication
between
them,
and
so
it's
actually.
This
is
actually
our
development
right.
Now,
let's
talk
a
bit
about
the
impact
and
what
we
did
in
our
organization.
A
So
each
machine
learning
model
that
we
will
built
manually
and
when
it
gets
into
our
platform,
it
got
improved
by
average
in
about
30%,
improved
by
improvement,
I
mean
performance
or
some
other
metrics
that
we
evaluate
model
by.
We
increase
the
number
of
machine
learning
models
in
about
70%
and
we
got
a
huge
growth
in
users
in
the
past
half
year
about
six
six
hundred
percent
of
world,
and
we
faced
today
a
lot
of
challenges.
One
of
them
is
the
code
remote
debugging,
not
every
data
scientist
walk
on
Jupiter.
C
We
actually
don't
want
to
keep
on
with
our
own
written
device
plug-in.
We
want
to
get
more
upstream
and
standardized
ID
device
plugin
that
everyone
can
use
and
to
share
the
GPU
as
we
want
for
multiple
pods.
As
we
explained
earlier
in
the
architecture,
and
we
have,
we
have
a
challenge
to
run
and
manage
the
multi
clusters
of
open
shifts
and
communities
to
get
manage
them
operate
and
monitor
a
lot
of
them
in
a
different
locations.
So
this
is
our
current
challenges.
B
So
that's
great
right:
did
you
guys,
like
that?
That's
great
right,
so
what
I
thought
I'd
do
here
is
kind
of
take
a
step
back
and
kind
of
see
what's
happening
right,
like
you
are
already
familiar
with
OpenShift
I
I
thought,
we'd
start
with
the
openshift
architecture.
You
can
see
your
master
and
your
worker
nodes
here
and
your
parts
running.
They
are
being
exposed
as
they're
being
exposed
as
services,
either
within
the
cluster
itself
or
through
the
routing
layer.
So
so
what's
and
then
all
of
this
you've
got
the
storage.
B
How
do
you
use
the
best
of
software
development
lifecycle
that
earlier
we
saw
with
Maguire
Bank,
for
example?
How
do
you
bring
that
to
basically
machine
learning
workflows
and
how
do
you
bring
that
into
production,
so
I
think
so
that's
kind
of
the
first
step.
The
the
next
step
really
is
some
of
the
stuff
that
it
I
showed
in
terms
of
you
know
the
Jupiter
notebooks,
for
example,
that
were
being
developed
that
he
showed
where
in
you
know,
it
can
spawn
off
multiple
workers
that
can
use
GPUs
etc.
B
Those
are
things
that
we
are
trying
to
that's
one
example,
but
then
there
was
also
an
example
of
how
workers
and
masters
communicate
with
each
other
using
somekind.
In
that
particular
case,
it
was
rabbit
MQ
there
was.
There
was
also
storage
that
he
showed
some
kind
of
interface
to
an
object:
storage
where
this
is
stored,
you
know,
so
what
we
are
doing
really
is
with
a
with
the
collaboration
internally
and
with
other
external
partners.
We
have
created
this
project
called
open
data
hub
and
open
data.
Hub
really
is
a
reference
architecture
built.
B
So
the
open
data
hub
has
two
aspects:
one
is
we
have
used
that
reference
architecture
internally
to
build
machine
learning
as
a
service
on
open
shift
at
Red
Hat,
and
we
are
using
it
internally
to
do
optimizations
for
our
data
scientists,
and
then
we
are
open
sourcing
it.
You
know
all
those
things
as
a
reference
architecture,
so
these
are
some
examples
of
those
things
that
are
shown
here.
There
is
like,
for
example,
Kafka
to
do
streaming.
There
is
an
endo
messaging
and
then
the
respond
to
do
a
streaming
in
other
data.
B
Real-Time
data
processing
there
is
the
Jupiter
itself
jupiter
hub,
wherein
it
has
this
pre-built
notebook
images,
and
you
can
add
more
as
you
need
you
know,
and
so
and
then
there
are.
We
have
a
ai
library
which
has
like
optimized
framework
such
as
tensorflow
built
on
there,
l
and
Red
Hat
stack,
which
includes
things
such
as
ubi,
which
is
Universal
base
image.
So
so
that's
basically
in
a
nutshell,
what
the
open
data
hub
is.
B
It
is
about
saying
that
if
it's
a
reference
architecture
that
you
can,
which
is
open,
source
based,
which
is
all
open
source
that
you
can
use
to
create
these
end-to-end
workflows
on
openshift
and
cuban
at
ease
for
building
ml
as
a
service,
deep
learning,
as
that's
one
aspect,
the
second
aspect
really
is
also
to
help
partners
right.
You
know,
you
know
if
there
is
a
partner
who
like,
for
example,
a
partner
such
as
anaconda,
for
example,
for
Jupiter.
B
You
know
they
could
use
this
reference
architecture
to
bring
their
products
and
services
on
top
of
OpenShift,
using
operators
and
operators
framework,
the
it
I
showed
Auto
ml
and
the
other
vendors,
such
as
h2o
driverless
AI,
which
can
do
the
same
thing.
So
so,
all
in
all,
we
are
excited
about
this.
We
welcome,
obviously
it's
open
source,
go
to
open
data
hub
IO
and
provide
pull,
requests
etc
and
provide
input
to
us
and
feedback.
B
The
other
thing
that
I
wanted
to
quickly
highlight
before
ending
really
is
our
continued
partnership
with
Nvidia,
which
around
this
stuff,
you
know
so
we
into
at
Summit.
We
introduced
this
idea,
a
program
called
accelerated
AI
and
what
it
is,
is
it's
really
a
easy
button
for
bringing
AI
and
creating
ml
as
a
service
in
the
enterprise
data
center.
B
You
know
libraries
for
machine
learning
so
that
program,
basically
so
what
you
get
really
with
this
so-called
with
the
accelerated
AI
program?
Is
you
get
the
NGC
containers
on
open
shift
on
x86
servers
with
GPUs,
fully
supported
end
to
end
from
Red
Hat,
obviously,
and
from
Nvidia
and
from
the
OEM?
So
that's
the
program
that
we
announced
we're
very
excited
about
that
there
is
a
lot
of
buying
at
both
companies,
as
well
as
the
OEM
level.
B
So
you
know
you
can
find
more
about
those
I
mean
there
are
those
resources
as
a
blog
on
it,
and
we
collaborated
with
in
video
on
that
and
then
we
are
also
inviting
customers
who
are
interested
in
participating
to
sign
up
if
they
are
interested
in
this
early
access
program.
You
know
so
anyway.
So
that's
where
I
wanted
to
conclude
and
thank
it
iron
guy
for
coming
over
and
talking,
and
we
are
very
excited
about
what
we're
going
to
do
with
respect
to
a
IML
on
OpenShift.
So
you
know
we'll
handle
questions,
offstage,
I.