Halodoc: Building the Future of Tele-Health One Microservice at a Time (Special Episode)

By | March 13, 2020


– Every successful business starts with a great idea. Today, moving from concept to reality, is easier than ever before. Join us, as we travel around the world to meet developers,
architects, and innovators, to learn how they’re
building on the AWS cloud. Welcome, to “This is my Architecture”. This is Jakarta, the capital of Indonesia. Located in the north west corner of the most populated island in the world. Rapid urbanization, coupled with competitive
investment in infrastructure, has made traffic pretty
intense around here. That means getting to a hospital
can sometimes take hours. And that’s just the travel time. Hospital queues can be so bad, some people make a living holding a place in line for others. In Indonesia there’s only two doctors for every ten thousand people. It’s no wonder most Indonesians only consider going to the
doctor when they’re really sick. So what if there was
a way to see a doctor, without having to get
onto these busy streets? Let me show you. – Hello Adrian, so I’m Doctor Nisa, show me how you feel today. – Well Doctor Nisa, I was, you know outside
today, its quite a warm day. And just outside (muffled speaking). This is Halodoc, a company and app, that allows you to interact with a doctor right from your phone. So, how is Halodoc good for your patients? – So you know, in Indonesia, there are many people lives
in rural areas, right, and they have restricted access
to health care facilities. Some would have to travel hours to get the the nearest
health care facilities. So, with Halodoc, you can just minimize the travel things, and then like in Jakarta you have traffic jams all
over the place, right? So you don’t have to travel, and you get to reduce the waiting time to get to the health care
facilities, and go to doctor. – That’s fantastic, thank you so much, doctor,
for helping me today. – Thank you Adrian,
hope you get well soon. – Thank you. – An app that makes it possible for a country of 264 million people, with 50% internet penetration, must have some serious
technology behind it. And now I’m going to
find out how it works. This is the nerve center at Halodoc, where right now dozens of
doctors are fielding inquiries from all around the country. As an island nation, 80% of calls are coming
from outside the city where access is much more difficult. – And to learn about
how all of this works, I’m joined by Abhilash, The Chief Technology Officer for Halodoc. Hi Abhilash, thanks for joining me today. – Thanks for having me. – So, I’m really keen to get underneath the covers of the architecture, but before we do that, tell us a bit about how it all started. – At Halodoc, we started
building our tech stack about three years ago. Our initial version was a simple monolith, one compute, backed by a simple database. – Wow, and obviously, things
have changed since then as you’ve continued to grow, so, walk us throw some of those changes. – Since our inception, Halodoc has started
launching new business lines, one of the business lines is
you can buy medicines online. Second business line is you
can talk to doctors online. As soon as we started
launching new business lines it was obvious to us a
monolith will not cut it and we had to go the micro service route as soon as possible. – So it sounds like, all these new interactions
that your users were having, became new micro services
in your architecture. – Each micro service has a
couple of compute instances, backed by assistance, we
use RDS MySQL database. So, one group could be
buying medicines online, the second group could be
talking to a doctor online. And so on and so forth, at the moment, we have 40 to 50 micro
services on production – So you’ve gone from using a
single class of EC2 instance to now having multiple different classes based on user interactions. Given that each of these micro services have their own databases,
I’m keen to understand, how did you think about data partitioning? – For certain use cases, we had to ensure that
the data is consistent across different micro services. A good example is, users’ data. As a user, you could buy
medicines or talk to a doctor, and these belong to
different micro services so what we did was we used DynamoDB to save all of our user data. – So I imagine, all of these services would
use identity quite often so having a low latency database, and being able to share that
across all these micro services is a big benefit. Now I see another service there, the simple notification
service, something that we see quite often in micro
services architectures. How are you using it? – So when we started our micro
service journey, I believe it was when we had about 10
micro services on production, we started noticing that we were having unnecessary tight coupling
between our micro services. One micro service would know
about the other micro service and would do a sync call. Right, so we wanted to kind of avoid that and bring in the asynchronous flexibility. So we essentially decided
to build an event bus, and that’s where the SNS journey started. – Now, you’ve obviously
gone from virtual machines to micro services. But I see you’ve gone even further, and you’re using server-less technology, so I’m keen to understand what
is Lambda doing over here? – Oh, lets take an example, lets say you spoke to a doctor, and doctor issued a electronic
prescription to you, right? So, the consultation service
would raise an event into SNS, and SNS would trigger Lambda. – Right, so you’re using Lambda to go and curate some of
the data in your event bus but, also to create
some stuff here as well. – And then, so, the Lambda
enriches the event raised, it pulls medicine names, the
doctor name, the patient name and then generates the PDF, which can be used by the
patient to buy medicines. – Fantastic, now what is this
second Lambda doing over here? – So one of the business use cases we had is real time visibility on
how certain metrics are doing. For example, availability of a doctor, how many consultations a doctor is doing, and so on and so forth. That was a requirement
from the business side. So the way we solved it is, each of these micro
services would raise events, it would go into SNS, from SNS we would send it to Lambda, Lambda would do the enriching
we spoke about earlier, and once the data is enriched
it would go into Kafka, and on top of Kafka we built Flink, then an elastic search, and
then Grafana for visualization. – The genius of Halodoc, was a merging of a network
of doctors and pharmacists. They connect 22000 doctors
and 1500 pharmacies, across 50 cities in southeast Asia. After a teleconference with a patient, doctors upload the medical prescriptions to the pharmacists, where they can be picked up and delivered. Good morning. – Good morning – So, Elisa how has Halodoc
changed your pharmacy business? – So how is it easier for your patients? – That’s great, so how does it all work? – So these are the Gojek drivers, they’re the ones that
pick up the medicines that patients need and deliver them right to their front door. And Teddy over here– – (speaks foreign language) – Is going to show me just how it’s done. Let’s go. – Okay. (high intensity percussion music) – These Gojeks are small,
swift, and very maneuverable. They can get potentially life
saving medication to patients in under 40 minutes. No wonder they’re so popular. And it’s all part of a
growing sector of the economy, ride sharing. – Another expanding sector
of the Indonesian economy, is eCommerce. Each has increased by 60%
over the last five years, and Halodoc is riding both those waves. So Abhilash, it’s clear that Halodoc really
deals with a lot of data, not just in volume, but also in type. Let’s start with what you do
with the transactional data. – [Abhilash] So from
all the individual RDSs, we are moving the data
into Redshift using DMS. – [Adrian] To the Database
Migration Service? – Yes. And on the Redshift, the
business analyst can run queries for monthly reports or weekly reports, and things like that. – Fantastic, obviously a very scalable
data warehouse solution. Now how do you keep that
RDS instance performant? – We want to ensure that
medical records are saved. We want to ensure that
historical record is kept safe, so the way we do it is from RDS older data gets moved into Glacier, for record keeping. – I imagine a lot of your analysts want to ask questions about
how the business is operating, tell me a little bit
about how you help them. – Good question, so lets say I’m an
analyst who is responsible for ensuring them there are
enough doctors on the platform. I want to know how many
doctors that are online, at the given moment. For this we built out
a real time pipeline, so the data from Kafka, gets ingested into Elastic Search, which acts as the persistent layer, and for the visualization, that’s what the analyst is going to use, we are going to use Kibana and Grafana. – I’ve noticed that you’re using a lot of open source software here, why is that? – Open source, there were two reasons one is the comfort developers
have with the tool, and second is the customizability, and the flexibility these tools offer. – When I was using the app, one of the things I really liked, was the different ways
I can actually interact. And one of those ways was I
could actually upload photos. Let’s dig into how you manage
that large image repository. – Sure, let’s say you’re
talking to a doctor, or a dermatologist in this
case, and you have a skin rash, so in the app you take
a photo of your skin, and upload it to the S3 bucket, right? So once the image is
uploaded to an S3 bucket, so it kicks in the image
recognition algorithm where we try to determine whether this is medically relevant or not. – So why actually filtering the images? – We want to ensure that
the doctors on the platform do not get inappropriate images from the users. – And Rekognition is
automatically trained to be able to remove potentially unsafe
images in the repository. – Yes, when we checked it on our data set, we found that we didn’t
have to retrain the model, because the Rekognition
did a really good job of classifying the images. – Health care is not a single product, but a series of interactions between patients, doctors,
pharmacists, and drivers. This was all previously done by hand. Halodoc had to create all the connecting tissue to make it work. – Abhilash, five million
users in just three years, you’re changing the face of
healthcare in this country, but I’m curious to understand what were some of those challenges that you’ve had to overcome along the way? – When we started building
technical solutions for business problems, we kind of had to challenge
a lot of assumptions. When we launched ability
to talk to a doctor online we kind of assumed that the data connectivity
would be solved for us. But when we actually launched, we were seeing significant
error rates up to 20%, but then we moved our stack to cloud, and to the MQTT protocol, after that we saw significant success. – What is some of the
other interesting areas that you want to take your product into? – Technically, two areas
that we are focusing on is on containerization and Kubernetes, and a lot of investment around AI/ML, because we believe that its going to kind of empower the whole ecosystem. We want to kind of personalize
our app and website to who the user is. To give one example we
have health content, we don’t want the same health content to be served to all users, for example, if you’re a diabetic, we want to serve this
diabetic content for you, and we believe that we can
achieve that using AI/ML. – What is the future for Halodoc? – At Halodoc we believe that we can use the power of technology to
change healthcare in Indonesia. We are working with Bill and
Melinda Gates Foundation, we have about 40000
midwives on the platform. We are building technology solutions to better enable midwives
to give infant care, in rural areas. – That’s fantastic, a big vision, not just for Halodoc, but also for this great country,
Abhilash thank you so much for letting me share
this day with your team, and also with yourself. – Thanks a lot. – I’ve learned a lot here in the last couple of days, and the citizens of
Indonesia are learning too. With access to healthcare in rural areas so difficult to access, and
with busy cities like this, Indonesians now have an option, Halodoc is changing things
with its convenience, and time savings. So, if people did more of this, they could avoid a lot of that, saving lives in the process.

Leave a Reply

Your email address will not be published. Required fields are marked *