Our AI is Cooler Than Your AI 

0:01 [Music]

0:07 the proliferation of devices data and

0:09 people have made it infrastructure more

0:12 complex than ever to manage

0:13 troubleshooting and information overload

0:15 threatens to slow down apps and

0:17 interfere with user experience

0:18 i'm rick roter from juniper a few days

0:20 ago i had the opportunity to sit down

0:22 with a few of my colleagues to discuss

0:24 ai and how it increases speed resolution

0:26 with the direct correlation user

0:28 experience let's take a look

0:31 gentlemen thanks for coming if you could

0:33 quickly introduce yourselves yeah rick

0:35 bob friday cto of juniper's enterprise

0:37 business

0:38 and thank you rick uh sudhir mata uh vp

0:40 of products here at the juniper's ai

0:42 driven enterprise right on well to kick

0:45 things off what are some of the concrete

0:46 areas that juniper and mist ai are doing

0:49 to address some of these challenges

0:50 you know i think you know there's

0:52 several good stories but i think one of

0:53 the most famous one for me was really a

0:55 big retailer where we were basically

0:57 trying to solve some little needle in

0:59 the haystack problem where they were

1:00 having these devices that couldn't

1:02 actually connect to the network randomly

1:04 coming going offline it was really a

1:06 good example where you know ai mutual

1:07 information was actually able to get to

1:09 the point of letting you know it was

1:11 really due to a particular os model type

1:13 that was having authentication problems

1:15 so that's really the power of ai is to

1:17 find these needle in the hey tax stack

1:19 problems that makes sense yeah and and

1:21 you know for us at juniper we measure

1:24 the impact of ai right so so there are

1:27 five things rick i tell every customer

1:29 you can take it to the bank if you come

1:31 to juniper you'll get these five things

1:34 the fastest deployment the fewest

1:37 tickets right imagine going into a

1:39 network and saying you know you know

1:41 service now as an example cut their

1:44 number of tickets coming into it

1:46 more than 90 fewer tickets right so

1:49 fastest deployment fewer tickets fastest

1:51 time to resolution any problem let's say

1:54 you know a doctor walks out of surgery

1:57 call drops now

1:59 and says hey why does my call drop

2:02 we're using ai and we're using machine

2:04 learning to go back and study sort of

2:06 recreate reality

2:08 and say what what really happened there

2:11 and what's all the you know the the

2:13 correlations and the causation of what

2:15 happened there compile all of that and

2:17 reduce the mean time to repair for for

2:20 solutions so first fast deployments

2:22 fewer tickets fast time to resolution

2:25 for tickets as they come in and then

2:27 last but not least is is dramatically

2:30 simplified operations these are the four

2:32 i.t outcomes you get in everywhere and

2:34 then bob's passion has always been is

2:37 how do we digitize the experiences for

2:39 employees guests patients passengers and

2:41 all these verticals and so we built a

2:44 platform for digitization and even in

2:47 that bob you've used a lot of ai even

2:49 the digitization yeah and this is really

2:51 around the location component

2:53 you know if you look at their

2:54 how are people connected to the internet

2:57 but when we get to our b2c customers

2:59 this is hospitality retail healthcare

3:01 wherever there is a consumer involved in

3:03 the business this is your b2c you know

3:05 what we really built as a platform that

3:07 allows you to basically engage with

3:09 those customers you know whether i'm at

3:10 a hospital i work with the hospital

3:12 really around uber for wheelchairs you

3:14 know like when a vet comes in right you

3:16 want to be able to find that uber they

3:18 want to get that wheelchair and get it

3:19 to the uber that requires location to

3:21 actually start working you know here at

3:23 juniper cova tracking during the covet

3:26 right you know you want to be able to

3:27 attract when people came in contact with

3:29 each other you know so that's the other

3:30 kind of use cases we're seeing around

3:32 this digital transformation of helping

3:34 our customers both be to see and even

3:36 carpeted enterprise customers who are

3:38 trying to facilities management now in

3:40 this new hybrid working space

3:42 that's amazing guys very cool stuff i

3:44 mean everyone is talking about ai ops

3:47 right so what do you think sets juniper

3:48 apart from the competition

3:51 well it's there there's no place in this

3:53 one right it's youtube guys right yeah

3:55 this one here is cloud and ai

3:58 right if you look at miss and what we're

4:00 doing now that differentiates us and

4:02 this is one reason why i left cisco to

4:04 start myth because i saw there was an

4:06 architectural opportunity that really

4:08 required a blank sheet of paper you know

4:10 that i wasn't going to be able to do at

4:11 a big company and it started building a

4:13 cloud that can actually do data science

4:16 and pipelines right

4:17 and that's basically solving that

4:19 problem around i'm not going to deliver

4:21 things every year i'm going to deliver

4:23 things every

4:24 week you know

4:26 as seder tells you christmas comes at

4:28 mist every week right you know every

4:30 week we get new features for our

4:31 customers the other big differentiation

4:33 is really around ai ops and this is

4:35 really that day two operational thing

4:38 and that's really what's changing the

4:39 industry we've seen ai ops change you

4:42 know how cars drive you know how doctors

4:44 diagnose cancer and image detection you

4:46 know in networking aiops is

4:48 fundamentally going to change how we

4:49 manage these networks of the future it's

4:51 really cool so this is a fundamental

4:54 question right

4:55 why what separates juniper mist rai ops

4:59 and ai platforms from any everybody else

5:01 in the industry right so tying off of

5:04 bob's uh answer there right bob said

5:06 it's cloud right so first you know it's

5:09 so important everybody has cloud meraki

5:11 has cloud aruba has cloud aerohive has

5:13 cloud everybody has cloud why is juniper

5:16 miscloud different it's as bob said it's

5:19 the pipeline to handle data on a per

5:22 user per minute basis when everybody

5:25 else's cloud

5:27 manages their network

5:28 we operate based on users experience so

5:32 every user every minute we're getting

5:34 that data into the cloud being able to

5:36 process that being able to you know

5:38 collect collate and curate that to sort

5:41 of make meaning on are we delivering a

5:43 great experience so it starts with the

5:45 cloud no question right and and what is

5:48 proof point very simple proof point bob

5:50 said you know uh like

5:52 when aruba central doesn't upgrade for

5:54 their cloud literally there's a 12-hour

5:57 outage on the cloud right you know the

5:59 the agility on some of our competitors

6:01 is terrible

6:03 christmas comes every wednesday at

6:04 juniper mist right literally we are

6:07 bringing new features to the cloud every

6:09 week every other week and this is

6:11 foundational agilities is important so

6:14 user experience every user every minute

6:16 and then an agile cloud that can

6:19 actually grow with the customers as

6:20 their need grows right and then so

6:23 that's on the cloud side on the platform

6:24 side fundamentally foundationally

6:26 different

6:27 yeah sorry bob you were great

6:29 i mean doing that wednesday christmas i

6:31 think is doing that with reliability and

6:33 so that's that micro service

6:34 architecture you know that's the

6:36 fundamental change going from an

6:37 embedded architecture that we were doing

6:39 at cisco 20 years ago to what we're

6:41 doing in the cloud that lets us do that

6:43 release every wednesday reliably right

6:45 without breaking anything speed and

6:47 flexibility flexibility with reliability

6:50 yeah and reliability

6:52 the one thing though on

6:53 when it comes to ai right bob is always

6:56 built as a pipeline it starts with data

6:58 it starts with you know great data you

7:01 know great grapes make good what good

7:03 wine comes from great grapes so i make a

7:05 barrel of wine every year you know great

7:07 wine starts with great

7:08 grapes

7:09 great ai starts with great data great

7:11 data right so great data for us and then

7:14 ai primitive framework that can take

7:16 that data and correlate and curate that

7:19 data and then comes complete a full

7:23 toolbox of algorithms we've been working

7:26 on for seven plus years and then

7:29 finally that all gets boiled up into a

7:31 nice little marvelous virtual assistant

7:34 we'll talk about marvis in a minute

7:36 but

7:36 all of this is a journey towards the

7:39 final answer the final answer for us is

7:42 a truly self-driving network right a

7:45 self-driving network meaning when you

7:48 deploy a juniper miss network i just

7:50 said we have fewer tickets how do we

7:52 have fewer tickets just because we threw

7:54 some ai algorithm that it no

7:56 we have the fewest tickets because we

7:59 use ai to design the network to deploy

8:02 and operate the network right channel

8:05 partners go in design they do site

8:07 surveys they deploy the network great

8:09 but the main the moment they walk off of

8:11 that premise that network is changing

8:13 the wi-fi is a living breathing thing so

8:16 we use reinforcement learning ai driven

8:18 radio resource management to sustain

8:21 that network at a strong point over a

8:23 long time right and i think the other

8:25 thing we've talked about is you know in

8:26 addition to that architecture is really

8:29 organizationally

8:30 what people don't appreciate with marvis

8:32 is in addition to the cloud architecture

8:35 is organizationally we have tied that

8:36 support team right to our data science

8:38 team and so people who are trying to

8:41 solve this problem that's probably the

8:42 other key secret sauce into marvis

8:45 is organizationally is making sure those

8:47 two groups are like acting as one

8:49 you know another thing that i hear a lot

8:51 about location data you brought it up uh

8:53 you know a little bit earlier but what

8:55 are some of the coolest ways that people

8:56 are using location data to to change the

8:59 way we live and work

9:01 yeah i mean so location is one of these

9:02 topics as dear seder knows has been dear

9:05 to my heart ever for the last 20 years

9:06 or so is really bringing indoor location

9:09 on par

9:10 with connectivity and making it a

9:12 must-have in the enterprise space you

9:14 know so what i said before is b2c

9:17 customers right retail after all these

9:19 customers are trying to digitally

9:21 transform either a customer experience

9:23 or an employee experience and we're

9:25 starting to see it on both levels you

9:26 know on the customer side it's about

9:28 wayfinding you know either i'm in a

9:30 complex hospital or a maze of things

9:32 they're trying to bring indoor

9:33 wayfinding on par gps

9:35 right you know the wheelchair example

9:37 right the uber right they're trying to

9:39 basically build applications you know to

9:41 help a doctor

9:42 can get a wheelchair to a uh a vet or

9:45 something right you know same thing you

9:47 want with your car right i want to be

9:48 able to sit there and i want to be able

9:49 to see you know how long am i waiting

9:51 are they coming they're making progress

9:53 that same experience you want indoors so

9:55 these are all examples that we're

9:56 starting to see

9:57 inside of our b2c customers and we're

9:59 also starting to see inside of our inner

10:01 carpet enterprise customers and ricky if

10:03 i could add a couple more interesting

10:05 examples right

10:06 april 2020 in march 17th uh 2020 so the

10:11 bay area here got shut down with covid

10:13 right

10:14 literally with under six weeks we turned

10:16 the juniper network right the same

10:18 network no new aps were added the same

10:21 network for complete digital contact

10:24 tracing right and so the the power of

10:27 the platform that sort of bob has

10:28 envisioned and built for us here is that

10:31 we are truly uh

10:33 truly today juniper networks is the only

10:36 networking player who's a leader in the

10:39 indoor location magic quadrant and so we

10:42 truly are the only player that can

10:44 combine best of breed wire and wireless

10:47 with this best of breed location

10:48 services so contact tracing especially

10:50 as come as employees are coming back

10:52 another very important use case as it's

10:55 so relevant now

10:56 is is during

10:59 and and even before that

11:01 companies were consolidating real estate

11:03 out right real estate is one of the

11:04 largest line items on on everybody's

11:07 uh you know expenditure and companies

11:09 are actually trying to optimize that and

11:12 how are they optimizing they're going to

11:13 digitize workspaces right so one of the

11:16 largest wins juniper we've had is a

11:18 financial services company and there

11:21 the the uh uh the facilities team

11:24 contributed a substantial money piece of

11:27 money for the infrastructure i.t project

11:29 because we're digitizing their employee

11:32 experience as employees are coming back

11:33 from kobe that's a remote working thing

11:35 right so we had a very large customer

11:37 that built a smart building you know

11:39 open workspace right hybrid workspace

11:41 hoteling right now but what they have to

11:43 figure out is why is no one going to

11:45 that side of the building you know is it

11:47 the temperature is it the furniture you

11:49 know so these are the type of problems

11:51 that we have to you know big companies

11:52 have to deal with when they move into

11:54 this new hybrid working model right now

11:56 making sure the space is being utilized

11:58 efficiently this is cool and it's almost

11:59 like even with iot i would imagine this

12:02 even becomes more important absolutely

12:04 absolutely

12:05 you know it's interesting

12:07 so so we talked a lot about digitization

12:09 location

12:10 but this conversation is about ai

12:12 how did we use ai for enabling this

12:14 digitization so let me give you sort of

12:16 a a very simple but very you know uh cut

12:21 and dry

12:22 you know comparison with our competitors

12:25 100 of our competitors every single one

12:27 of them if they need to deploy location

12:30 they need to go like uh the old verizon

12:32 ad can you hear me now can you hear me

12:34 now they need to go to every point in

12:36 the building and do manual calibration

12:40 bob you've used ai for that so explain

12:42 that yeah so at miss actually our very

12:45 first patent

12:46 got issued in less than six months was

12:48 for virtual ble you know in the very

12:50 first use case for aiml was basically

12:52 trying to learn the path loss model in a

12:54 building for every device every model

12:57 inside the building so that was actually

12:59 our first example of aiml was really

13:01 around location and what that really

13:04 solved is the problem of making sure

13:05 people didn't have to do fingerprinting

13:07 and a lot of setup you know because when

13:09 you look at location the biggest problem

13:11 with it has been the complexity of

13:13 getting it up and running it's an

13:14 overlay network

13:15 right if you look what miss has done

13:18 every access point we ship has a virtual

13:20 bla antenna built in already so that

13:23 totally eliminates the overlay problem

13:25 right which is what seder is pointing

13:26 out yeah you know the cloud is what let

13:28 us get to you know in six weeks we had

13:29 covet up and running covert tracking up

13:31 and running that was the cloud

13:33 but the fact that we had virtual ble was

13:35 the other part of it that everything was

13:36 enabled for location coming out of the

13:38 bat coming out of the gate you put that

13:40 in context if i'm a retailer let's say i

13:42 have a thousand retail stores

13:44 if i choose if they choose our

13:46 competitor they choose an aruba cisco

13:49 meraki

13:50 they actually have to send people

13:52 trained people to a thousand stores to

13:55 calibrate them right millions of dollars

13:58 versus a juniper network that is using

14:00 ai for calibration think of applying ai

14:03 in a dramatically different place well

14:05 in that case the virtual bla eliminated

14:07 the overlay aiml eliminated the

14:10 calibration factor and so that was we

14:13 were the first ones to basically get rid

14:14 of the need to calibrate these networks

14:16 for location

14:17 that's crazy it's crazy it is

14:20 what a great opportunity to hear from

14:22 sudhir and bob to learn about juniper's

14:24 ai capabilities and hear how great ai

14:27 starts with great data thank you for

14:29 joining and be sure to check out all the

14:31 great content that we have about ai on

14:33 the feed