►
From YouTube: Applied ML weekly team meeting Aug 5, 2021
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
A
So
first
time,
let's
remind
ourselves
to
use
the
stream
live
for
youtube
features.
So,
let's,
let's
continue
the
discussion
on
kafka,
open
source
versus
cockpit,
commercial
via
confluent
versus
google,
pub
7.
A
for
the
proof
of
concept
and
post
the
frequent
concept
the
where
alexander
and
mihao-
and
I
left
it,
I
believe-
is
that
we
we
need
more
time
to
really
evaluate
to
see
if
we
want
to
use
a
commercial
solution
for
pub
sub
like
confluent,
so
we've
already
integrated
kafka,
which
is
in
the
commercial
version,
but
with
a
developer
license
and
as
long
as
we
can,
which
is
free.
As
long
as
we
can
stand,
the
developer
onenote
license
for
the
proof
of
concept.
A
I
think
that's
great
for
the
proof
of
concept
and
then
post
proof
of
concept
we
can
consider
do
we
want
to
continue
with
that
and
pay
for
it.
Do
we
want
to
go
open
source
and
not
pay
for
it,
but
have
more
work
for
us
to
do
and
we're
learning
is
for
us
to
do
or
use
google
pub
sub
instead,
which
is
not
free.
We
pay
for
that
too,
but
it's
part
of
our
existing
gcp
contract.
What
are
your,
what
are
your
thoughts
on
there.
B
I
think
for
the
for
the
proof
of
concept,
it's
fine
right
now
to
use
confluent
platform
with
the
development
license.
It's
understood
right
now,
if
everything
is
fine
after
our
last
meeting
with
with
confluence
right
in
general,
I
think
that
maybe
we
don't
need
to
have
kafka
on
the
customer's
side.
So
in
this
case,
if
we
don't
need
kafka
there,
so
we
can.
We
can
stay
with
kafka
in
our
own
google
cloud
instance
or
we
can
switch
to
pops
up.
B
B
It's
not
like
beam
apache
beam
and
python,
sdk
and
kafka.
So,
as
I
said,
and
I'll
in
in
the
in
in
one
of
the
issues,
so
it's
not
that
easy
to
connect
python
sdk
for
beam
and
kafka.
B
So
that's
why,
for
now
I
decided
to
connect
data
beam
or
data
flow
in
our
case,
with
postgres
directly
without
using
kafka.
A
Okay,
I
didn't
catch
all
of
you,
but
if
you
could
add
your
nose
to
the
the
doc
that'd
be
great,
let
me
do
it
now,
of
course,
that'd
be
great
for
them.
So
just
a
quick
question:
are
we
using
any
features
or
configurations
in
fluid
for
the
poc
that
would
require
us
to
use
a
trial?
License
would
be
free
for
30
days
versus
the
developer,
license
to
be
free
forever.
B
B
Yeah
we
have
this
connector
to
move
data
from
kafka
to
google
cloud
storage,
but,
as
I
understood,
if
we
use,
if
we
use
one
broker
and
this
connector,
that's
still
the
development
license.
B
A
Can
you,
okay,
I'm
talking
to
them
again
later
today,
yeah
and
I.
A
B
Oh
yeah
change
so
yeah
so
right
now
I'm
working
connecting
data
flow
and
postgres
the
connection
between
the
extract
stage
and
transform
stage
is
done.
So
we
need
to
to
make
the
last
connection
we
need
to.
B
We
need
to
to
change
the
to
change,
to
change
the
code
inside
each
of
the
component,
because
we
change
the
database
because
before
we
use
manga
right
now
we
just
postgres.
So
that's
why
we
need
to
change
some
of
the
things.
Then
we
need
to
to
run
this
pipeline
at
this
time.
We
need
to
write
to
run
it
manually
because
we
don't
have
any
airflow
or
we
don't
connect
to
to
to
gitlab
pipeline.
B
So
we
need
to
run
it
manually
and
then
so
once
we
once
we
get
the
data,
we
can
train
the
model,
put
it
on
the
put
it
inside
kubernetes
and
finally,
we
need
to
update
the
user
interface
so
check
what
maybe
we
maybe
we
will
break
something,
but
I
hope
no
and
then
so
and
that's
all
so.
I
hope
yes,
that
will
finish
until
the
end
of
this
month,
but
maybe
like.
B
I
would
like
to
have
at
least
one
week
more
just
to
be
to
be
sure
that
everything
is
fine,
but
I
hope
yeah,
I
hope
yeah.
I
will
try
to
do
to
do
the
best
to
finish
everything
during
this
month.
B
A
A
B
So
I
like
how
we
work
with
the
mighty,
because,
like
I
rolled
the
yamas
young
definitions
to
provision
postgres
and
kafk
and
after
that
might
wrote
terraform
scripts
to
to
make
it
to
make
it
better.
So
so
we
split
the
work.
A
B
Yeah
we
need
to
we
need
to.
We
need
to
do
the
same
thing
with
terraform
and
kafka
scientists
too,
because
this
issue
is
not
closed.
Right
now
but
yeah,
I
wrote
all
the
yaml
definitions,
all
the
yaml
files
with
the
documentation
how
to
run
them.
So
I
hope
that
we
will
close
this
issue
soon.
A
Jumping
back
a
bit,
can
you
slap
me
later
with
the
link
or
put
in
the
note
and
or
put
in
the
notes
on
the
plugin,
the
adapter
that
the
rights
from
the
confluent
connector
that
we're
using
so
I
can?
I
can
review.
B
You
mean
that
we
use
right
now.
Yes,
this
is
the
described
one.
C
Okay,
so
we
should
welcome
eduardo
bonnet,
who
was
hired
by
bartek
as
our
ml
ops,
full
stack
engineer
in
incubation
engineering,
david
and
I
have
talked
through.
How
will
square
incubation
engineering's
call
to
action
with
what
this
work
with
this
group
is
doing,
and
ultimately,
our
main
thought
initially
here
is
that
if
there
are
things
that
are
broken
or
could
work
better
with
running
data,
science,
workloads
within
gitlab
ci
eduardo
will
be
in
a
good
position
to
be
able
to
fix
those
things
for
us.
C
So
I
know
like
there
are
existing
bugs
with
our
jupiter
notebook
integration
that,
I
think,
would
be
some
easy
first
issues
for
eduardo,
but
as
we
start
building
this
out,
I
think
if
you
have
ideas
or
things
that,
like
I
just
really
wish
it
did
x,
that
will
be
a
good
focus
for
eduardo.
Initially
we're
still
trying
to
square
really
what
incubation
engineering
is
and
how
it
relates
to
the
model
ops
stage
more
to
come
on
that
I'm
actually
meeting
with
eduardo
here
in
15
minutes
and
kind
of
gonna
walk
him
through.
C
Where
we've
been,
where
we're
going
what's
next,
so
yeah
feel
free
to
to
grab
coffee,
chats
and
say
hi.
Consider
him
kind
of
part
of
this
group
he's
kind
of
off
on
his
own
little
island
very
similar
to
the
way
we
are
here.
So
I
think
we
will
find
a
lot
of
synergy
with
him.
He
is
based
out
of
amsterdam,
so
alexander,
I
think
he'll
overlap
more
with
your
day,
so
hopefully
that'll
be
someone
that
will
end
up
being
much
more
of
a
counterpart
to
you.
A
Yeah,
I'm
doing
one
with
him.
I
think
it's
tomorrow
and
I
definitely
encourage
others
as
well
to
make
it
work.
Also,
I
did
add
an
issue
for
milestone
to
events,
to
get
what
eduardo's
been
working
on
and
it
is,
is
already
working
on
an
ml
offset
potential
use
of
it
in
the
applied
ml
team.
I'm
sure
some
of
it
will
apply
and
I'm
sure
some
of
it
will
at
least.
B
D
D
Thank
you.
Sorry,
I'm
mostly
yeah,
I'm
mostly
learning,
so
I'd
rather
listen
rather
than
talk.
Thank
you
very
much
alexander
taylor
and
mikhail
for
the
work.