►
From YouTube: Behind the Scenes at OpenShift Commons: Interview with Joshua Hoffman(Bloomon) by Micheal Hausenblas
Description
(Joshua Hoffman, Bloomon)
Joshua Hoffman, Bloomon
A
B
Very
much
yeah
and
these
these
lovely
flowers
are
a
big
part
of
what
we're
doing
here.
So
I
am
the
CTO
at
lumen
and
we
are
a
flower
subscription
company,
but
we're
actually
also
a
tech
company,
and
you
might
be
similar
to
me
when
I
first
heard
of
bloom
on
I
thought,
flowers,
tech,
how
you
know,
maybe
you're
collecting
an
address
and
sending
it
to
a
flower
shop
and
there's
a
delivery.
B
You
know,
what
do
you
need
a
tech
team
for,
but
actually
we
like
to
say
bloom
on
is
where
code
turns
into
flowers,
and
technology
is
a
part
of
everything
that
we
do
so,
to
give
you
a
little
bit
of
background,
we're
based
in
Amsterdam,
which
is
in
many
ways
the
epicenter
of
the
floral
universe,
and
you
can
50%
of
the
world's
flowers,
for
example,
either
originate
here
or
pass
through
here
now.
This
has
been
happening
for
a
long
time
and
for
50
to
100
years.
B
There
has
been
no
new
technology
introduced
to
this
industry,
and
so
what
that
means
is
you
have
a
lot
of
people
involved,
a
lot
of
traders
that
are
at
multiple
layers
of
the
business,
a
lot
of
stuff
done
by
hand
and
a
lot
of
really
inefficient
stuff.
My
favorite
example
is
that
you
have
tulips.
They
are
grown
in
Kenya.
They
are
brought
to
Amsterdam
to
an
auction.
B
They
are
sold
to
a
local
buyer,
who
will
then
turn
around
and
sell
them
to
a
distributor
that
covers
Japan,
who
will
take
them
possibly
to
Tokyo,
sell
them
to
a
local
distributor,
who
will
then
eventually
bring
them
to
a
flower
shop
and
with
a
perishable
product?
You
know
you
can
imagine,
the
lifespan
is
really
cut
down,
and
typically
the
flowers
are
out
of
the
ground
seven
to
ten
days
before
they
end
up
in
a
flower
shop
and
in
your
hands.
B
I
was
quite
surprised
by
this
as
well.
So
the
short
version
of
what
we're
doing
here
is
we
wanted
to
accomplish
two
things:
one.
We
want
to
make
a
really
fantastic
product
available,
and
part
of
that
is,
of
course,
the
lovely
designs
that
our
designers
come
up
with
and
incorporating,
perhaps
unusual
plants
and
flowers
you
wouldn't
find
in
a
flower
shop.
But,
more
importantly,
we
want
to
apply
technology
to
this
and
through
the
application
of
our
technology
that
we've
developed
and,
of
course,
strategic
relationships.
A
So
essentially,
it
needed
not
show
what
you're
doing
there
in
terms
of
supply
chain,
optimization
whatever
you
want
to
call
that
that's
applicable
to
other
industries
as
well,
but
it's
not
just
the
flower
industry,
but
you
are
kind
of
pioneering
that
there
and
applying
it
there
and
yeah,
potentially,
because
a
lot
of
people
wake
up
here.
Right,
yeah.
B
Or
I
was
a
flower,
absolutely
I
mean
we
say.
Our
company
mission
is
to
spread
happiness,
and
this
comes
in
just
like
you
said,
so.
We
of
course
have
supply
chain
technology
that
involves
some
basic
stuff,
like
digitizing
parts
of
the
supply
chain
that
were
written
by
hand
before
so.
You
know,
growers
typically
did
not
have
any
kind
of
relationship
or
connection
to
the
flower
shops.
Instead,
they
would
all
meet
up
at
the
auction
house,
bring
their
products
in,
and
you
know
hope
to
get
the
best
price
and
that's
it.
B
B
It
turns
out
that
the
way
most
pookas
are
assembled
is
again
a
very
manual
thing.
You
know,
flower
shops
are
fairly
small.
They
have
a
few
workers
there,
they're
assembling
beautiful
bouquets.
How
do
you
do
this
at
scale?
And,
of
course,
if
you're
like
me,
you
say
what
we
need
technology
and
so
what
we've
done
is
in
addition
to,
of
course,
all
the
software's
that
we
develop.
That
allows
us
to
have
these
relationships
with
growers.
We
also
have
builds
and
I
call
it
a
semi,
automated
production
line.
We
have
a
proprietary
machine.
A
B
Absolutely
probably
one
close
to
your
heart:
we
use
kubernetes
to
host
our
production
environment,
and
this
of
course
allows
you
know
all
the
great
benefits
that
you
get
there.
Our
teams
can
easily
deploy.
We
can
easily
manage
stuff
pretty
much.
We
are
entirely
cloud
which
is
for
me
an
interesting
transition.
So
this
is
the
first
company
I've
worked
out
where
the
infrastructure
in
terms
of
software
and
technology
is
100%
in
the
cloud,
and
we
also
have
this
big
physical
operation.
B
So
it's
a
little
bit
of
bulk,
which
is
a
lot
of
fun,
so
projects
that
we
work
on
that
I
can
share
with
you,
for
example,
on
this
production
machine
I
should
say
also
so
we're
really
passionate
about
metrics
metrics
drive
everything,
and
this
means
that
in
some
cases
we
are
working
to
physically
instrument
parts
of
the
production
facility,
adding
buttons
using
cameras
to
try
to
observe
events
that
are
happening.
All
kinds
of
fun
stuff
like
that,
and
eventually
all
flows
back
into
a
metrics
platform
and
our
favorite
metrics
platform
shouldn't
surprise.
A
B
Indeed,
you
know,
and
the
thing
is
you
know
what
a
flowers
care
about:
will
they
care
about
the
environment,
take
care
about
the
temperature,
the
humidity?
These
are
all
things
we
can
collect
and
graph
and
understand,
and
we
even
have
a
lab
where
in
tightly
controlled
situations,
we
put
each
of
the
weeks
or
months
bouquet
and
we
monitor
their
health
in
this
humidity
and
this
temperature
and
that
humidity
and
that
temperature-
and
we
record
it
on
video
with
pictures.
B
A
B
So
we're
just
getting
into
that
now,
it's
very
early
days,
but
so
the
majority
of
our
purchasing
is
designer
driven.
We
have
a
gentleman
named
Anton,
who
is
a
flower
artist,
and
he
comes
up
with
these
wonderful
concepts
and
collections
for
our
bouquets,
but
we
are
collecting
all
the
time
information.
We
call
things
like
last
week's
flower
rating
and
what
we
want
to
understand
from
all
of
our
customers
when
they
receive
the
flowers
they
will
rate
it,
but
not
just
on.
Did
you
like
it?
But
what
did
you
think
of
the
colors
that
were
used?
B
B
What's
popular
now
and
what's
likely
to
be
popular
for
the
next
season,
he
can
incorporate
this
into
his
designs
and
then
we
can
augment
what
we
send
to
the
growers
to
say:
hey
we're
going
to
you
know,
need
these
these
and
those
and
in
these
colors
and
using
machine
learning,
get
better
and
better
at
predicting
what
people
will
like
it's
a
really
interesting
challenge
in
that
the
typical
consumer.
So
this
is
a
luxury
product
right.
Flowers
are
a
wonderful
addition
to
your
home,
but
no
one,
you
know
starves.
B
If
you
don't
have
a
flower,
it
turns
out
that
the
people
that
love
our
product,
the
most
the
subscribers
they
don't
want
to
say
what
they
want.
They
want
to
be
pleasantly
surprised,
and
so
this,
for
me,
is
a
great
opportunity
for
machine
learning,
and
so
this
is
where
we're
exploring
now
to
bring
this
technology
and
say:
ok.
Well,
we
sent
Michael
a
bouquet
every
week.
These
are
the
ways
he
scored
these
various
bouquets.
We
have
classified
in
detail.
A
Right
and
actually
think
about
it,
it
also
has
a
kind
of
like
so
even
like
social
media
expected,
like
you
know
at
the
app
here
saying
like.
Oh,
you
know,
people
who
upset
their
better
half
in
the
same
way
that
you
do.
You
know
typically
pick
that
right.
So
is
that
also
kind
of
like
the
plan
that
you
know
not
just
great
stuff,
but
you
kind
of
like
get
these
recommendations
based
on.
Maybe
your
interactions
or
your
you
know,
connections
or
whatever.
That's
you
yeah.
B
Long
term
we're
we're
always
hungry
for
data,
always
looking
for
more
input,
so
the
the
short-term
vision
where
that
starts
is
we
say
something
like
you
know.
Christmas
is
coming
up
based
on
your
tastes
and
the
ratings
that
have
come
from
the
gifts
you've,
given
your
family
in
the
past,
we
think
here's
three
things
that
would
make
a
lovely
Christmas
gift.
You
know
pick
the
one
that
makes
the
most
sense
to
you
or
just
go
with
our
recommendation.
I
do
think
it
it.
B
You
know
it's
an
interesting
future
to
think
about
where
you
know
my
mother
loves
flowers
and
I
frequently
send
them
to
her.
For
occasions
and
I'm,
not
sure
you
know
the
the
the
part
that
I
find
interesting,
I'm
very
comfortable
having
the
software
tell
me
a
recommendation
that
she
would
like,
and
especially
when
it's
proven
itself
right,
multiple
times
and
I
get
credit
right
like
she
doesn't
think
about
where
I
came
from
I
wonder
what
do
we
go?
B
B
I'm
very
much
the
same
and
I
think
that
this
sort
of
element,
also
of
a
delightful
surprise,
is
something
that
you
know.
I
cannot
ask
you
in
advance.
What
would
you
like?
Is
it
surprise
for
your
birthday,
because
then
there
is
no
surprise
so
having
a
way
to
sort
of
understand
a
bit
about
your
tastes
and
and
then
come
up
with
a
delightful
surprise.
That,
to
me
is
the
magic.
A
B
C
A
A
B
We
had
I
mean
for
me,
so
some
of
the
really
basic
lessons
you
know
there's
a
difference
between,
of
course,
logs
and
metrics,
and
when
you
get
into
data
processing
it
becomes
a
whole
nother
set
of
issues.
So,
for
example,
and
if
I
miss
five
minutes
of
metrics,
we
can
move
on,
we
can
live
with
that.
We
have
a
lot
of
data
to
look
at
if
something
goes
wrong
in
our
data
pipeline,
where
we're
actually
building
metrics
about
the
business
and
the
flowers
and
so
forth.
This
is
a
much
more
complex
issue.
B
When
I
first
arrived
here,
we
had
situations
where,
in
the
very
early
tech
stack,
there
were
jobs
that
would
run
before
we
adopted
airflow,
and
if
you
didn't
run
that
job
in
the
same
15-minute
window
every
day,
you
could
not
reproduce
the
results.
So,
if
anything
went
wrong,
you
didn't
get
the
same
result
anymore
running
it
an
hour
later,
and
this
is
a
great
way
to
start,
but
it
doesn't
take
you
very
far.
B
So
what
we
learned
and
what
we
get
out
of
air
flow
is
this
coordination
and
really
having
the
dependency
graph
understood
by
the
platform
itself.
So
if
something
goes
wrong,
we
can
restart
just
the
things
that
need
to
be
restarted
and
make
sure
they're
restarted
in
the
right
order.
We
can
parallel
eyes
as
much
as
possible
and
avoid
parallelizing
the
things
that
shouldn't
do
and,
most
importantly,
by
making
sure
that
our
data
team
is
well
disciplined
and
that
we
have
really
strict
standards
around
our
data
warehouse.
B
Everything
that
we
process
can
be
reproduced
and
you
will
get
the
same
result,
and
that
is
critical
feature
and
not
even
something
that
I
thought
about.
You
know
in
my
career
up
until
now,
just
how
critical
that
reproducibility
is,
and
you
look
at
a
lot
of
external
analytics
platforms
like
you
know:
Google
Analytics,
it's
fast
and
easy
to
set
up
and
wonderful,
but
you
have
a
four-hour
window.
A
B
I
think
probably
computer
vision,
driven
machine
learning.
It's
really
fascinating
to
me
because,
as
I
work
to
just
capture
metrics
from
the
operations
in
the
physical
plant
in
the
production
facility,
you
know
sometimes
it's
very
easy.
You
can
put
a
small
sensor
here
and
capture
temperature.
You
can
attach
an
Arduino
to
some
machine
and
every
time
the
button
is
hit,
you
can
read
that
button
press,
but
there
are
other
things
happening
and
there
are
efficiencies
and
the
way
everything
moves
throughout
the
facility
and
for
us,
as
we
scale
up
as
a
business.
B
We're
really
seeing
these
efficiencies
become
more
and
more
important,
so
because
our
ambition
is
always
to
have
a
same-day
turnaround,
so
the
flowers
come
fresh
in
the
morning
from
the
growers
and
they
need
to
be
processed
so
becoming
more
and
more
efficient.
One
of
the
things
we're
exploring
is
putting
cameras
in
strategic
places
and
using
computer
vision
to
identify
and
learn
from
the
patterns
of
movement
within
the
facilities
and
also
the
various
events
that
might
occur.
B
B
But
it's
important
to
dream
big
is
being
able
to
look
at
the
flowers
as
they
grow
and
understand
their
state
of
health,
and
so
we
see
a
future
potentially
where
you
know
we
send
out
drones
to
fly
over
the
fields
and
take
observations
of
the
state
of
things
growing
and
feed
that
back
into
our
system
as
well.
You
know
really
the
sky's.
The
limit
I
think
you
know
with
an
industry
that
is
hungry
for
technology
and
hasn't
had
it
right
in
years.
B
B
A
B
B
No,
it's
just
a
name
that
our
founders
came
up
with,
so
you
have
the
Dutch
word:
BL,
o
ye
M
bloom
yeah,
but
our
name
is
not
spelled
this
way.
It's
BL
o
om,
o
n,
and
so
it's
sort
of
like
a
combination
of
bloom
and
monitoring,
because
we
want
to
be
the
high
tech
solution
of
flowers
by
foreign
customers.
It
just
means
beautiful
flowers,
good,
but
I
xx
bear
yeah.
We
are
bloom
on
and.