Download as pdf or txt
Download as pdf or txt
You are on page 1of 38

MongoDB.

local

Partner with MongoDB and AWS to


offer Online-Merge-Offline service

Mariana ChowAlan Chan


Solutions Architect
Founder & CEO of Omnichat
Amazon Web Services
(AWS) Mariana Chow
Solutions Architect, Amazon Web Services (AWS)
Agenda
● MongoDB Atlas on AWS Reference
Architecture
● Case Study: Onmichat’s Online-Merge-
Offline service
● Summary and Next Step
MongoDB Atlas on AWS
Modernize Your App with Atlas and AWS AppSync

AWS AppSync AWS Lambda MongoDB Atlas

User
Modernize Your App with Atlas and AWS AppSync

AWS AppSync AWS Lambda MongoDB Atlas

Amazon Bedrock

User
Amazon Bedrock
Choice of foundation models

NEW NEW NEW

JURASSIC-2 COMMAND + EMBED STABLE DIFFUSION XL 1.0 CLAUDE 2 AMAZON TITAN


Multilingual LLMs Text generation Generation of unique, LLM for thoughtful Text summarization,
for text generation model for business realistic, high-quality dialogue, content creation, generation,
in Spanish, French, applications and images, art, logos, complex reasoning, classification, open-
German, embeddings model and designs creativity, and coding, ended Q&A,
Portuguese, Italian, for search, clustering, based on Constitutional AI information extraction,
and Dutch or classification in and harmlessness training embeddings and
100+ languages search
Amazon EC2 P5 instances Amazon EC2 P4d/P4de instances Amazon EC2 G5 instances

Powered by NVIDIA H100 Powered by NVIDIA A100 Powered by NVIDIA A10G


Tensor Core GPUs Tensor Core GPUs Tensor Core GPUs

Up to 6x faster and up to 40% Up to 2.5x faster and up to 60% Up to 3.3x higher performance
lower training costs than previous generation
cost-to-train savings than previous G4dn instances
generation than previous generation
GPU-based instances P3/P3dn instances
Purpose-built accelerators
for generative AI

AWS Trainium AWS Inferentia2


Up to 50% savings on training costs Up to 40% better price performance
over comparable Amazon EC2 instances than comparable Amazon EC2
instances
GENERALLY AVAILABLE Generate code suggestions
in real-time

Amazon Scan code for hard-to-find

CodeWhisperer vulnerabilities

Build apps faster and more Flag code that resembles


securely with an AI coding open-source training data or
filter by default
companion
FREE FOR INDIVIDUAL TIER
Analyze Your Data in Atlas

AWS Glue
Data Catalog

MongoDB Atlas AWS Glue Amazon S3 Amazon Amazon


Athena QuickSights

User
Build Your Event-driven App with Atlas and EventBridge

Other System

MongoDB Atlas Amazon AWS Lambda


EventBridge

User
Orchestrate Your Workflow using AWS Step Functions

MongoDB Atlas Amazon EventBridge


Alan Chan

Partnering with
Founder & CEO of Omnichat

MongoDB and AWS to


offer Online-Merge- Mariana Chow
Solutions Architect of
Offline Services Amazon Web Services (AWS)
Agenda
Introduction of AWS & Omnichat
Use Case
Data Usage
Challenges faced
Solutions & Result (MongoDB Atlas)
Solutions & Result (AWS)
Milestones
●Meta Business Partner, WhatsApp BSP
●Launched Marketing Cloud ●Entry into SEA Market
●Established in Hong Kong ●Whatsapp Integration & ●Headcount: 100+ HK, TW, SG
●Launched Customer Service Official Partnership
Cloud
100+ staff across APAC

2017-2018 2019 2020 2021 2022 2023

●Entered Taiwan Market ●Launched OMO Sales Cloud ●Launched Social CDP
●Integrated LINE & Facebook ●Instagram Integration ●Launched Omni AI (ChatGPT)
Chatbot ●Official LINE Technical Partner ●Launched WhatsApp Payment
●WhatsApp Global Select Partner
●LINE Silver Technical Partner
Presence
Trusted by
South
Korea 5,000+ companies
Hong
Kong
Japan

Thailand
Taiwan

Philippines

Indonesia

Malaysia
Australia
Singapore
NGOs
# Our Presence
Key clients
Education & Hobbies Food & Groceries
Fashion Health & Beauty

Others
Watch & Jewelry

Home Appliance
& Electronics
Partnership & Supporting Platforms
E-Commerce platform Messaging platform Platform Partner
*Official partner provider

Others
Retail & Marketing Awards
(2022-2023)
Meta success story

https://developers.facebook.co https://business.whatsapp.co https://business.whatsapp.co


m/success-stories/sasa-hk/ m/resources/success- m/resources/success-
stories/venchi stories/benefit-cosmetics
Omnichat Chat Commerce Products
Omnichannel CRM Marketing Online-merge-Offline
Messaging Automation Sales
OFFLINE Salesperson B
Customer

Consolidate all the social messengers Identify the customers interest by


into one single platform tagging and deliver automatic
(IG/FB/WeChat/WhatsApp/Line) marketing Drive online merge offline sales and
Enhance remarketing and boost trace back the online contribution of
conversion: cart abandonment brick-and-mortar salesperson
Utilise chatbot to answer basic reminder, coupon and game module
enquiries and facilitate auto-tagging Build up conversational workflows with
Deliver a customised shopping sales & CS reports for efficient
segmentation
experience in all channel (Web-push/ management monitoring
Broadcast via Facebook
Messenger/WhatsApp)
Segmentation Broadcast
FB Ad -> Unique OMO Coupon
15 mins, 2000 coupons 2 hours, 4000 coupons
FB/IG Click to WhatsApp Ad -> OMO Coupon
Chatbot
Land on Select
FB/IG Receive Redeem
WhatsApp eCoupon
WhatsApp Game for acquisition

Playing games in Win a coupon as


WhatsApp interact to trigger start WhatsApp an award
Playing games
Online

MTR ad for acquisition


Land on WhatsApp

MTR Ads

Offline
Taxi body ad for acquisition
WhatsApp eCoupon
Offline Taxi
Latest development – Omnichat Social
CDP
Marketing Automation message for VIP
Purchased
XOVE Skincare

After 1 Month

If customer is VIP, auto


send VIP specific discount
message to users
If not VIP, will not send any
message
Abandoned Cart Retargeting Journey
Start the converting customer journey
after users add item to Shopping cart

Auto send 1st


personalised
messages to
After 2 Hours audience to convert
users with product
added to cart
without purchase

Auto send 2nd


personalised
After 1 Day messages to
audience to convert
users with product
added to cart
without purchase
Data Usage
Customer Profile Customer Behavior Data Messages
(~34M documents) (~1.6B documents (~400M documents)
quarterly)

- Build up customer persona data - Keep track on all customer


- Support complex marketing conversations
automation criteria based on - Easy to read in single platform
customer attributes, tagging & & searchable
behavior

- Understand the customer via messaging


- Achieve accurate targeting to increase
conversion
Challenges faced
1. Unstructured data
- Customer attributes & tagging vary among different businesses
- Limited fields supported in customer profile data / message types due to the
defined data model
- Not easy to alter the data model when the data size growth

2. Scalability
- Data size growth exponentially when our business growth
- Performance issue on large data size

3. Efficient & low latency search functions on messages


- Synchronization issues
- Cost on maintenance on the data pipeline
- Data unsync issues when the data pipeline is downed
Solutions & Result
(Migrate data to MongoDB Atlas)

1. Customer data
- Benefit from the document database nature of MongoDB
- Able to store complex & unstructured data for different businesses
- Easier for developers to build features on the flexible document data model

2. Scalability
- Support horizontal scaling via sharding
Solutions & Result
(Migrate data to MongoDB Atlas)
3. Build message search function using Atlas Search
- No data pipeline is required anymore
- Reduce on maintenance cost
- Low learning curve on rebuilding the search query using Atlas Search
- Migrate to Atlas search within 2 weeks

4. Current Usage on Atlas


- 3 clusters (M40/M50) for customer profile, messages, customer behavior events
data
- Data size
- Customer profile: ~150GB
- Messages: ~200GB
- Customer behavior data: ~600GB
Solutions & Result
(AWS)
Starts with a simple infrastructure
• Route 53 as DNS Service to route user
requests to our applications
• Application Load Balancer to distribute
user traffic across our application servers
hosted on EC2 instances
• RDS as database
• S3 for file storage

• Whole infrastructure is managed by AWS


• Focused on application development
Solutions & Result
(AWS)
• During annual holiday season, special
sales or new product launches

• AWS Auto Scaling Group

- Dynamic scaling based on instance metrics


or on a schedule
- No need to plan ahead
- Simple setup yet effective
- Cost effective (auto scale up & down)
Next Step: Together with AWS and MongoDB
Competencies/Programs Joint Integrations

AWS PrivateLink AWS Wavelength

Amazon EventBridge Amazon Kinesis Data Firehose

• AWS Marketplace Seller AWS Key Management Amazon Simple Storage Service
• Financial Services Competency Service
• AWS PrivateLink Ready Product AWS Identity and Access
AWS Lambda
• AWS Public Sector Partner Management
• Govt. Software Competency
Amazon Managed Streaming for
• Data & Analytics Software Competency AWS IAM Identity Center Apache Kafka
Thank you
Alan Chan Mariana Chow
Founder & CEO Solutions Architect
Omnichat Amazon Web Services (AWS)

You might also like