►
From YouTube: CDF SIG MLOps Meeting 2020-07-16a
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
B
A
A
A
B
Sorry
ya
got
a
problem
there
with
a
microphone
yeah.
So
it
looks
like
come.
It's
just
gonna,
be
us
this
week,
so
I
don't
think
we
can
pursue
the
usual
agenda.
But
if,
if
you'd
like
to
spend
a
few
minutes
talking
about
your
interests
in
in
in
the
roadmap-
and
you
know
and
the
thing
that
you
have
questions
about,
then
you
know
now's
a
good
time
to
to
jump
in
yeah.
A
Project
inside
drinking
sex,
I
guess
I'm
just
trying
to
like
see
what
people
are
trying
to
talk
and
then
eventually
maybe
contribute
some
time
I
mean
I
have
been
contributing
to
drinking
sex
and
the
docs
and
telephone
stuff.
For
some
time
now,
but
yeah
I
mean
I
need
to
learn
probably
more
about
cute
flow
and
stuff
like
that.
B
So
yes,
well,
what
we're
trying
to
do
here
is
to
is
to
provide
a
sort
of
generic
view
of
what
the
requirements
are
from
an
ml
ops
perspective.
You
know
what
the
what
the
challenges
really
are
and
what
what
that
implies
for
technical
requirements
and
and
potential
solutions,
and
and
then
you
know
off
the
back
of
that
there.
There
are
then
multiple
candidate
solution
projects
that
are
addressing
some
or
all
of
the
concerns
within
the
roadmap.
A
But
are
there
any
yeah
I
mean?
Are
there
any
yeah
tickets
in
the
JX
yeah,
a
repo
which
deal
with
ops,
I
thought
I
saw
some
what
did
not
really,
but
now
that
I
check
for
it
I
don't
see
any.
B
So
so
there
there
are
one
or
two
things
outstanding
at
the
moment:
the
the
key
piece
being
that
we
need
to
work
on
the
the
versioning
of
things
like
builders
and
build
packs,
so
that
it
makes
it
easier
for
us
to
support
machine
learning
projects
in
the
in
the
longer
term.
So
so,
right
now
with
in
Jenkins
X,
this
there's
a
kind
of
Big
Bang
approach
to
things
like
quick
starts.
B
So
you
can
end
up
in
a
situation
at
the
moment
where
you're
you're
sitting
on
a
particular
version
of
a
Jenkins,
X
development,
environment
and
creating
projects
based
on
quick
starts
and
you
you
can
be
building
something
in
one
way
on
one
week
and
then
the
following
week.
You
try
to
do
the
same
thing
again
and
you
get
a
completely
different
build
pipeline
because
the
the
quick
starts
been
updated
and
the
build
packs
been
updated
and
you're
you're
getting
a
later
version.
B
A
B
And
there
is
some
very
some
prototype
versioning
code,
I
think
in
the
QuickStart,
but
it
wasn't
implemented
so
far
as
I
recall.
So
I
think
there
is
a
there's,
a
command
option
to
actually
select
a
version,
but
it
doesn't
appear
to
do
anything
at
the
moment.
So
that
might
be
a
place
to
have
a
look
and
see
yeah
see
what
that
allows
you
to
do.
A
B
B
We're,
rather
than
pursuing
the
normal
agenda,
where
we're
just
having
a
bit
of
a
free-for-all
in
terms
of
answering
questions
and
generally
exploring
so
yeah
feel
free
to
you
know,
have
a
chat
about
what
your
interests
are
and
how
you'd
like
to
get
involved
in
in.
In
what
we're
doing
and
ask
me
anything
you
like
about
about
the
project
and
what
we're
trying
to
put
in
place.
Yeah.
C
A
C
Right
now,
I
really
find
you
know
like
the
world
of
ml
and
Emma
Lopes,
very,
very
interesting
and
I'm
looking.
You
know
how
to
help
this
community
grow
and
maybe
how
to
create
some
innovation
in
the
area.
So
right
now,
I'm
just
about
you
know
contributing,
because
when
you
give
you,
you
almost
always
get
something
back
on
the
community.
So
that's
my
agenda.
Okay,.
B
B
We're
still
a
very
early
stage
of
maturity
in
terms
of
everyone,
understanding
what
the
challenges
are
and
having
a
clear
view
of
the
problems
that
they're
solving.
So
the
the
idea,
certainly
this
year
is,
is
to
just
get
the
road
map
pulled
together
so
that
we've
got
a
communication
document
that
we
can
say,
look
here's.
C
C
You
know
it's
remind
me,
the
days
that
you
used
to
log
into
the
FTP
server
and
modify
the
files
in
all
agates.
They
are
not
like
a
good
working
habits
right
now
in
the
industry,
everything
that
regarding
2ml
and
I,
really
like
your
initiative,
because
it's
pushed
the
water
industry
into
you,
know
more
professionalized
working
other
than
experimenting
called
experimenting.
It's
that's
something
you're
doing
the
university
in
the
lab,
but
in
the
real
world
you
need
to
have
a
walking
standards.
C
B
And
I
think
this
is
doubly
important
for
for
AI,
because
you're
typically,
the
types
of
problems
that
you're
building
area
products
to
solve
are
going
to
be
decision
making
products
rather
than
decision
support
products
and
and
therefore
you're,
actually,
betting.
The
farm
on
the
output
of
this
piece
of
software
and,
and
so
the
standards
that
will
be
expected
of
these
products
will
actually
be
considerably
higher
than
we
expect
from
our
conventional
software
products.
And
yet
the
the
techniques
and
best
practices
are
not
established
to
that
level.
B
So
the
the
more
that
we
can.
We
can
mentor
and
educate
the
data
science
group
and
then
support
them
with
the
tooling
that
they
need
to
get
their
work
built
to
it
to
the
relevant
quality.
So
it
can
be
deployed,
managed
node,
the
more
will
accelerate.
You
know
AI
in
general
as
a
practice,
so
so
there.
So
the
roadmap
is,
if
you
like,
the
the
requirements
gathering
piece
and
at
this
stage,
where
we're
focusing
mostly
on
on
the
generic
requirements.
B
B
So
on
on
that
project,
we're
basically
trying
to
align
to
the
road
map
and
and
build
a
CI
CD
platform
that
will
implement
the
the
requirements
that
we
are
setting
out
in
in
the
road
map
and
you'll
find
that
there's
a
there's
a
first
release
of
that
available
now.
So
there
is
a
a
soft
release
of
some
early
capabilities
within
Jenkins
X,
but
there's
still
a
lot
of
work
to
be
done
there
to
cover
off
all
the
problems
that
we
we
can
see
as
challenges
in
the
road
map.
So.
C
B
B
C
B
The
semiconductor
roadmap
he
is
currently
focusing
on
effectively
inference
per
watt
as
a
metric,
because
we
need
to
be
reducing
the
the
power
cost
of
inferencing
to
negligible
levels
so
snow
that
we
can.
We
can
push
that
intelligence
out
into
the
network
and
so
you're
you're
going
to
be
deploying
your
models,
probably
into
silicon
in
tiny
devices
that
are
distributed
in
their
billions
across
the
edge
of
the
network.
B
Yeah,
so
the
problem
is
going
to
be
really
interesting
because
it's
it's
not
going
to
be
the
simple
thing
of
oh:
let's
just
make
a
model
in
Python
and
then
stick
it
in
the
cloud.
It's
gonna
be:
oh,
let's
make
a
model
in
in
some
abstraction
layer
and
then
turn
that
into
silicon
and
right
through
foundry
and
then
ship
the
devices
all
over
the
world,
and
so.
B
C
B
The
the
real
thing-
the
real
thing
to
think
about
here
is
that
the
you
know
there
was
that
cheesy
phrase
a
few
years
about
the
you
know,
the
network
is
the
computer,
but
that
is
now
truly
being
implemented
as
that
as
the
new
default.
So
so,
instead
of
thinking
about
what
infrastructure
am
I
deploying
on
to
in
a
data
center?
C
To
a
building
a
little
bit
going
back
because
you
know
a
few
years
ago,
most
of
the
software
was
deployed
on
the
edge
edge
devices
and
then
it's
Swift
to
their
to
the
cloud
and
now
you're
saying
because
we
need
low
latency,
maybe
instead
of
deploying
edge
servers
like
edge
clouding,
we
need
to
go
a
little
bit
back,
because
this
way
that's
the
only
way
it
can
be
fast
interesting.
So.
B
C
B
B
So
so
the
the
challenge
that
we
face
going
forwards
is
how
do
we?
How
do
we
have
the
commercial
arrangements
to
allow
us
to
to
train
a
piece
of
IP
and
deploy
it
in
a
location
where
it
can
be
executed
and
trusted
to
run
against
a
particular
set
of
data
when
all
of
these
things
are
owned
by
different
individuals,
and
you
know
managed
independently
from
a
commercial
perspective.
C
B
So
so
they're
there
there
are
a
number
of
known
issues
and
we're
we're
trying
to
capture
some
of
them
and
they're
in
the
roadmap.
You
know,
typically,
if
you're
working
with
say,
let's
look
at
a
chemical
process,
control
scenario
you
you,
you
can
easily
today
cover
your
entire
process
line
in
sensors
that
give
you
lots
and
lots
of
valuable
data
about
what's
actually
happening
within
that
chemical
process
line.
B
C
B
Then
somewhere
along
line
you've
got
to
execute
all
of
this
stuff,
so
you've
got
to
actually
add
a
push
code
back
into
the
system
and
have
it
modify
the
the
state
of
the
equipment
of
particular
points
on
the
line.
So
you
go
then
I
also
have
a
trust
relationship
to
be
able
to
execute
code
effectively
inside
somebody
else's
IP
stack,
so
so
we're
we're
facing
a
an
inversion
of
architectures
from
a
security
privacy,
confidentiality,
intellectual
property
perspective,
which
is
probably
a
bigger
problem
than
actually
solving
the
AI
and
machine
learning.
One
yeah.
B
C
Interesting,
so
maybe
I'll
tell
you
a
little
bit
about
how
how
Canaria
you
know
so,
except
of
you
know,
I
I
can
help
also
in
the
technical
stuff,
but
I
think
that
I
make
you
settle,
but
then
I'm
really
good
with
communities
and
people
I
opened
the
just.
A
few
days
ago,
three
meetups
Emma
Lopes,
Tel,
Aviv
ml
of
San
Francisco
and
emma
loves
New
York,
and
you
know,
because
we
are
in
the
corner
time.
We
can
actually
it's
easier
because
you
can
just
create
a
zoom
it
up
and
I.
C
C
B
That's
that's
explicitly
what
the
roadmap
is
about.
It's
about
saying.
You
know,
here's
here's!
What
is
here
today,
here's
what
we
expect
to
be
coming
over
the
next
five
to
ten
years,
and
he
is
a
view
of
you
know
what
the
current
day
is
versus,
where
we
need
to
be,
and
therefore
you
can
see
where
the
gaps
are
and
where
there
are
commercial
opportunities
to
be
had
in
investing
in
building
product
in
that
area
or
where
there
are
societal
benefits
in
investing
in
in
in
building
open
source
to
to
solve
those
problems.
I.
C
B
C
There
and
you
know
it's
win-win
situation
and
help
the
community
this
way
I.
If
we,
if
we
gonna,
you
know,
use
this
meetup
for
the
group
goals.
Maybe
we
need
to
think
about
what
we
want
to
achieve
from
this
meetup
that
will
benefit
the
group,
so
maybe
I'll
chat
with
you
about
it.
You
know
offline
and
slack
and
write.
B
C
We
want
to
ask
the
community
some
questions.
Maybe
we
want
some
presentation
that
talks
about
the
stuff
we
want.
So
that's
about
that
and
the
second
thing
is
feel
free
to
you
know
Jimmy.
You
know.
If
you
want
something
to
be
done,
don't
don't
sometimes
you
know
I
I
just
ignore
stuff,
because
I
want
them
to
be
on
read
so
I'll
can
and
I'm
gonna
handle
them
later.
So
just
feel
free
to
announce
me
I'm
volunteering
myself.
So
it's
always
good
to
have
someone.
You
know
doing
that
the
boring
stuff
for
you
so
I'm.
B
No
problem
well
likewise,
if,
if
you're
interested
in
joining
any
of
the
implementation
projects,
then
you
know
there
were
plenty
of
people
who
will
be
pitching
you
on
on
helping
out
on
on
on
their
projects.
Yes,
certainly
more
than
welcome
to
come
and
play
in
the
Jenkins
xbase
if
you're
interested
you
know.
Likewise,
there
are
a
lot
of
other
projects
under
this
cdf
banner
that
are
all
looking
for
people
to
muscle
in
and
move
them.
C
Forward
only
what
I'm,
what
I'm
planning
to
do
is
to
find
some
project
that
I
think
it's
interesting.
You
know
personally
and
I
can
somehow
be
with
it
to
some.
You
know
among
some
sometime,
I
will
need
money,
so
I
need
to
start
my
startup
soon.
So
I'll
try
to
find
how
can
I
combine
my
contribution
to
the
community
in
a
way
that
I
can
build
a
company
with
it,
because
that's
my
job
and
by.
B
C
C
C
B
B
All
right,
well,
I,
think
I
think
we're
probably
best
to
close
this
one
day
and
early
today,
given
that
we
we
don't
have
a
large
attendance
this
week,
but
what
I'll
do
is
make
sure
that
I
stay
in
touch
with
you
and
do.
B
B
C
B
You
yeah
well,
we
try
and
keep
these
fair,
fairly
informal,
and
you
know
so
this.
This
is
an
opportunity
for
for
people
to
get
together
and
actually
dive
into
some
of
the
challenges
in
this
space
and
they're.
Typically,
what
we'll
do
is
we'll
we'll
have
a
chat
about
a
particular
challenge
in
in
one
of
these
sessions
and
then
we'll
go
away
and
write
that
up
and
incorporate
it
into
the
document.