Redis Case Studies iFood's Utilization of Redis Cloud for Enhanced Machine Learning Operations

Edit This Case Study Record

	iFood's Utilization of Redis Cloud for Enhanced Machine Learning Operations Redis

iFood's Utilization of Redis Cloud for Enhanced Machine Learning Operations

Redis

Technology Category	Analytics & Modeling - Machine Learning Platform as a Service (PaaS) - Application Development Platforms
Applicable Industries	E-Commerce Retail
Applicable Functions	Logistics & Transportation Procurement
Use Cases	Last Mile Delivery Retail Store Automation
Services	Cloud Planning, Design & Implementation Services Training
Challenge	iFood, a popular food ordering and delivery service in Brazil and Colombia, faced a significant challenge in maintaining the performance of its machine learning (ML) models. The company's success was directly tied to the performance of these models, which needed to process data quickly to reduce costs, increase revenue, and influence user behavior during real-time interactions. The COVID-19 pandemic presented unique opportunities for e-commerce firms, especially online delivery services that were prepared to handle an escalating volume of orders. At iFood, the technology team had to manage millions of new users and thousands of new restaurants joining its platform. Despite the surge in business volume, iFood remained committed to providing an optimal experience for its customers. Read More
About Customer	iFood is an online food ordering and delivery service based in São Paulo, Brazil. The company maintains an 80% share of the Brazilian food delivery market and is also widely used throughout Colombia. iFood serves a network of about 320,000 restaurants and aims to increase its base of 30,000 supermarkets. It also recently added delivery for pharmacies and liquor stores. The company uses artificial intelligence (AI) and machine learning to better understand users, such as tracking how many orders they have placed in the last month, which restaurants and stores they prefer, their chosen payment mechanisms, and many other variables. Read More
Solution	To ensure consistency for ML engineers and create an optimal experience for users, iFood uses Redis Cloud on AWS as the foundation of its rapidly evolving ML feature store. Redis on Flash maximizes data-processing throughput while reducing overall data storage costs. Redis Cloud makes feature data available to dozens of models in production, and it includes essential capabilities such as a registry, data pipeline, and monitoring tools to streamline feature engineering activities. This allows iFood’s Data & AI team to search, reuse, and serve features in production, at scale. Redis worked closely with iFood to implement Redis on Flash and integrate the fully managed service into iFood’s microservices architecture. The key to this exceptional performance is the way Redis on Flash orchestrates data between the DRAM and Flash Storage tiers. Read More Log in to view content
Contents

Technology Category

Analytics & Modeling - Machine Learning

Platform as a Service (PaaS) - Application Development Platforms

Applicable Industries

E-Commerce

Retail

Applicable Functions

Logistics & Transportation

Procurement

Use Cases

Last Mile Delivery

Retail Store Automation

Services

Cloud Planning, Design & Implementation Services

Training

Challenge

iFood, a popular food ordering and delivery service in Brazil and Colombia, faced a significant challenge in maintaining the performance of its machine learning (ML) models. The company's success was directly tied to the performance of these models, which needed to process data quickly to reduce costs, increase revenue, and influence user behavior during real-time interactions. The COVID-19 pandemic presented unique opportunities for e-commerce firms, especially online delivery services that were prepared to handle an escalating volume of orders. At iFood, the technology team had to manage millions of new users and thousands of new restaurants joining its platform. Despite the surge in business volume, iFood remained committed to providing an optimal experience for its customers.

About Customer

iFood is an online food ordering and delivery service based in São Paulo, Brazil. The company maintains an 80% share of the Brazilian food delivery market and is also widely used throughout Colombia. iFood serves a network of about 320,000 restaurants and aims to increase its base of 30,000 supermarkets. It also recently added delivery for pharmacies and liquor stores. The company uses artificial intelligence (AI) and machine learning to better understand users, such as tracking how many orders they have placed in the last month, which restaurants and stores they prefer, their chosen payment mechanisms, and many other variables.

Solution

To ensure consistency for ML engineers and create an optimal experience for users, iFood uses Redis Cloud on AWS as the foundation of its rapidly evolving ML feature store. Redis on Flash maximizes data-processing throughput while reducing overall data storage costs. Redis Cloud makes feature data available to dozens of models in production, and it includes essential capabilities such as a registry, data pipeline, and monitoring tools to streamline feature engineering activities. This allows iFood’s Data & AI team to search, reuse, and serve features in production, at scale. Redis worked closely with iFood to implement Redis on Flash and integrate the fully managed service into iFood’s microservices architecture. The key to this exceptional performance is the way Redis on Flash orchestrates data between the DRAM and Flash Storage tiers.

Impact #1	iFood's Redis-based infrastructure delivers sub-millisecond performance for ML operations, powering a highly efficient microservices architecture that optimizes the user experience, allows for massive scalability, and enables rapid corporate growth. The company's ML models help bring in revenue, such as the ones that power its recommendation engine and determine which types of vouchers and coupons to present to users. Other models help iFood reduce costs by detecting fraudulent transactions. The use of Redis Cloud has also encouraged reusability and collaboration among the team, preventing the need to 'invent the wheel' multiple times. The exceptional stability and reliability of Redis Cloud, which utilizes a shared-nothing cluster architecture to automate failover operations at the process level, for individual nodes, and across infrastructure availability zones, have also been beneficial for iFood.

Impact #1

iFood's Redis-based infrastructure delivers sub-millisecond performance for ML operations, powering a highly efficient microservices architecture that optimizes the user experience, allows for massive scalability, and enables rapid corporate growth. The company's ML models help bring in revenue, such as the ones that power its recommendation engine and determine which types of vouchers and coupons to present to users. Other models help iFood reduce costs by detecting fraudulent transactions. The use of Redis Cloud has also encouraged reusability and collaboration among the team, preventing the need to 'invent the wheel' multiple times. The exceptional stability and reliability of Redis Cloud, which utilizes a shared-nothing cluster architecture to automate failover operations at the process level, for individual nodes, and across infrastructure availability zones, have also been beneficial for iFood.

Benefit #1	In the first three months of 2021 alone, more than 30,000 new restaurants, grocery stores, and convenience stores signed up to deliver food through iFood’s digital platform.
Benefit #2	iFood’s monthly orders surged by tens of millions of deliveries.
Benefit #3	Redis Cloud offers less than one millisecond per Read operation, compared to Amazon DynamoDB's 10 milliseconds per Read operation.

Benefit #1

In the first three months of 2021 alone, more than 30,000 new restaurants, grocery stores, and convenience stores signed up to deliver food through iFood’s digital platform.

Benefit #2

iFood’s monthly orders surged by tens of millions of deliveries.

Benefit #3

Redis Cloud offers less than one millisecond per Read operation, compared to Amazon DynamoDB's 10 milliseconds per Read operation.

Download PDF Version

Overview

iFood's Utilization of Redis Cloud for Enhanced Machine Learning Operations

Operational Impact

Quantitative Benefit