►
From YouTube: Applied ML weekly team meeting July 1, 2021
B
Worries
so
we
continue
to
make
great
progress.
We
have
a
bunch
of
people
from
other
teams
here
today,
which
is
great
that
are
learning
about
what
we're
up
to
I'll
just
give
a
quick
and
alexander.
I
just
got
a
message.
He
said
he'll,
be
here
in
a
minute:
just
got
a
quick
update
for
for
matt
and
sean
and
alper.
B
Is
that
and
sean
knows?
Some
of
this
is
a
new
team
we
just
founded.
Recently.
We
are
in
the
process
of
integrating
the
underview
technology,
which
is
going
well
so
far
in
it
for
a
proof
of
concept.
So
we've
got
in
staff
for
milestone
one
and
we're
gonna.
Do
it
for
git
lab
itself,
first
to
dog
food
it
ourselves
before
we
fully
integrate
it
in
the
future
and
so
far
so
good
mihawk.
You
wanna
give
a
quick
update
on
what
you've
been
working
on.
B
A
B
D
Yeah,
I
think
it
looks
like
an
infrastructure
request.
I
don't
know
the
structure,
but
for
product
intelligence
we
had
a
gcp
project
with
specific
permissions,
but
anytime
you
work
with
the
infrastructure.
I
think
you
have
to
expect
several
days
of
delays
because
they
are
usually
overloaded
with
other
issues.
So
I
think
that's
expectable,
but
in
case
anything
I
can
help
it.
D
I
have
a
gcp
project
for
product
intelligence,
which
is
not
used,
but
I
think
we
shouldn't
use
it
for
that
purpose
anyway,
and
I
don't
think
there
are,
there
will
be
permissions
for
kubernetes
there
already.
So
that
will
be
another
question
mark.
E
One
thought
on
this:
is
it's
good
that
we're
trying
we're
dog
fooding
first
and
that
you
know
the
alderman
customer
will
be
gitlab
employees.
So
if
there
are
problems,
you
know,
that's
actually
not
a
big
issue
for
our
customers,
but
I
guess
this
may
be
something
that's
quite
new
to
infrastructure
and
sre,
and
perhaps
we
would
like
to
you
know
have
some
I
wouldn't
say
training,
but
some
type
of
you
know
education
or
what
they
might
need
to
do.
E
If
there's
a
problem
and
and
this
this
team
can't
be
reached
for
whatever
reason
like.
E
Interact
with
either
infrastructure
or
sre
and
kind
of,
let
them
know
that
we're
going
to
have
this
other
cluster,
it's
going
to
be
separate.
You
know
how
it
would
be
managed
things
like.
B
B
Taylor,
you
want
to
do
your
quick
update
on
what
you've
been
working
on.
Are
you
blocked?
What
are
you
going
to
work
on
next,
etc?.
F
Sure
so
getting
those
private
milestones
we
talked
about
yesterday,
set
up
with
all
of
the
check
boxes
and
get
that
organization
and
issue
trees
set
up.
I'm
trying
to
wrap
my
head
around
all
of
the
nyi
headcounts
and
how
we're
going
to
adjust
the
organization
of
the
model
ops
stage
to
take
on
anti-abuse,
which
I
guess,
I'm
the
default
product
manager
for
and
then
direction
page
updates
to
reconcile
a
lot
of
the
things
I
previously
mentioned.
C
Yeah,
sorry,
I'm
a
bit
lost
today
I
was
replacing
hyphen
adf
with
data
flow
and
I
lost
myself
a
bit
yeah.
So
right
now,
as
I
said,
I'm
working
on
replacing
azure
data
factory
and
hive
with
the
data
flow.
So
I
finish
almost
everything,
but
I
write
the
code
in
jupyter
notebooks,
so
I
have
to
move.
I
have
to
make
the
code
production
ready
for
the
right
now,
so
we
are
able
to
read
data
from
the
local
files,
but
generally
we
need
to
read
them
from
kafka.
So
that's
will.
C
C
Yeah,
sorry,
I
was
trying
to
say
that
once
we
install
the
kubernetes
environment,
for
instance,
if
you,
if
you
need
my
help
to
install
kafka,
I
can
do
that.
B
A
Basically,
the
idea
is
that
I
can
either
create
a
new
project
within
the
within
the
gcp
organization
that
we
have
and
put
things
in
that
project,
or
I
could
try
adding
a
new
kubernetes
cluster
to
an
existing
project
which
could
be
gitlab
internal
or
there
are
several
projects
available.
But
I
am
not
sure
that
would
be
the
correct
approach
and
I
basically
rather
make
sure,
with
the
infra
team,
that
this
is
what
we
are
doing
actually
makes
any
sense
or
I'm
just
introducing
problems.
B
Thanks
thanks
for
the
detail
of
it,
sean
you
mentioned
earlier
about
giving
the
infrastructure
team
a
heads
up.
We
did
actually
give
them
a
heads
up
on
the
proof
of
concept
the
proof
of
con,
and
they
gave
us
advice
on
how
to
plan
for
post
with
the
concepts.
The
proof
of
concept
is
which
we're
talking
through
this
to
get
feedback
on
it.
B
We
are
pulling
data
from
git
lab
itself,
the
gitlab
project
itself,
using
existing
apis,
pulling
it
into
the
machine,
learning
components,
processing
the
data
and
then
likely
are
going
to
be
passing
the
results
back
to
the
machine
learning
as
a
comment
or
maybe
even
display
it
in
a
separate
ui.
We
haven't
decided
yet
so
it's
all
separate
and
it's
using
existing
functionality
and
it's
on
one
project
past
the
proof
of
concept.
B
It's
actually
going
to
be
a
feature
that
we're
going
to
make
available
inside
the
product
itself
for
uh.com
customers
and
that's
where
we're
not
going
to
be
pulling
the
api.
We
think
that
will
be
too
slow
and
particularly
load
on
the
system.
We're
going
to
pull
directly
from
the
database
and
it'll
be
directly
integrated
with
the
product
so
there'll,
be.
You
know,
portions
of
the
ui
on
the
product
itself.
That's
what
will
help
get
help
and
advice
from
matt
and
the
front
end
team
for
verify
and
others
to
make.
B
E
B
Data
flow,
of
course,
the
whole
unreview
code
base,
which
is
a
large
amount
of
large
percentage
of
it,
is
python
which
there
isn't
much
of
any
python
in
gitlab,
but
that's
the
de
facto
standard
for
learning
and
a
number
of
other
things
as
well
that
are
brand
new.
We
did
eliminate
a
number
of
new
things
like
we're
not
going
to
use
mongodb.
That's
part
of
the
current,
the
previous
architecture.
B
C
Yeah,
so
I
have
been
contacted
by.
Let
me
check-
oh
sorry,
so
by
the
security
automation
team
and,
as
I
understood
right
now,
they
work
on
entities
on
the
anti-spam
service.
So
if
we
I
can
send
the
link
right
now.
So
if
we
check
the
architecture,
it
looks
like
they're
also
trying
to
integrate
the
something
like
kafka.
C
So
maybe
so
they
send
me
a
link
to
schedule
a
coffee
chat.
I
can
talk
with
them.
Maybe
we
can
find
something
similar,
maybe
the
way
how
they
are
trying
to.
They
also
have
some
machine
learning
models
and
maybe
in
the
way,
maybe
the
way
how
they're
trying
to
learn
how
they're
trying
to
train
these
models
and
how
they're
trying
to
aggregate
data.
Maybe
it
looks
similar
to
what
we
are
doing
right
now.
B
I'm
just
gonna
do
a
little
bit
of
the
heading
gate
level.
That's
awesome
for
transparency.
I
don't
know
how
they
found
out
about
it.
Maybe
it's
the
handbook
updates
that
you've
been
doing
on
the
architecture,
but
that's
great.
I
would
recommend
an
additional
coffee
chat.
You
know
maybe
recording
it
and
putting
it
unfiltered.
B
C
I
put
link
into
the
doc
file
on
the
issue
that
they
send
me.
So
if
we
check
their
architecture,
I
mean
the
picture.
We
can
see
that
there
is
a
message
broker.
I
don't
know
what
they
mean
by
this
message
broker,
but
maybe
kafka
can
be
used
also,
and
if
we
look
at
this
architecture,
it
looks
similar
to
what
we
are
trying
to
achieve
in
the
milestone
too.
C
B
C
B
A
A
B
C
Well,
sorry,
going
back
to
the
first
point:
should
we
check
the
handbook
page
right
now,
it's
written
here,
updating
the
private
milestones
check
marks.
So
in
this
sense,
do
we
do
do
we
need
to
check
the
handbook
page.
B
Yeah,
maybe
maybe
probably
good
for
taylor
to
take
since
you're
updating
the
what
could
be
confidential
versus
not
in
the
issues
and
ethics.
Take
a
look
at
the
handbook
pages.
If
you
would
taylor,
make
sure
they're
consistent
too.
B
Yeah,
just
real
quick
before
you
end,
I
just
want
to
introduce
myself
so
I'm
the
backend
engineering
manager
for
the
code
review
team.
So
at
some
point
I
assume
we'll
be
working
closely
together,
so
just
wanted
to
at
least
chime
in
and
say
hi
and
say
nice
to
meet
you
and
looking
forward
to
working
with
you.