Rippling - System Design Interview

Staff Software Engineer

Problem Statement

Rippling develops products for HR domain. Design a system to support the following

1. Upload documents.

2. Update the verification status of the document.

3. Fetch the verification status of the document.

Volume

1. 500K documents to be processed in 24 hours.

SLA

1. Each documents needs to be processed within an SLA of 1 to 4 hours depending on the priority.

Assumptions

1. Assume the documents are verified manually by an operations team.

2. Each document is associated to an employee with 1:1 relationship.

Duration

60 minutes on CodePair

---

Solution

APIs

Define APIs with request and response body structures.

1. /createDocument

API used to upload a new document. Needs to have validation to verify if document of same type already exists for the given employee and organisation combination. The default verification status as part of the API will be unverified

HTTP Method : POST

Payload :

Request:
{ employeeId : xxx, 
organisationId : yyy, 
documentInfo : {    // Extensible Json payload associated to document such as binary payload, documentId, documentType etc } 
} 
Response :
{ 
employeeId : xxx, 
organisationId : yyy, 
documentId :  // unique Identifier for the document 
}

2. /updateDocument

Used to update document verification status.

HTTP Method : PUT

Payload :

Request:
{ employeeId : xxx, 
organisationId : yyy, 
documentInfo : {    // Extensible Json payload associated to document such as binary payload, documentId, documentType etc } 
} 
Response :
{ 
employeeId : xxx, 
organisationId : yyy, 
documentId :  // unique Identifier for the document 
}

3. /getDocumentStatus

Used to fetch verification status of the document

HTTP Method : GET

Payload :

Request:
{ employeeId : xxx, 
organisationId : yyy, 
documentId :  // unique Identifier for the document 
 
} 
Response :
{ 
 verificationStatus : // valid values
}

Database

There’s no one correct solution and we can make it work with either choice of SQL or NoSQL database with combination of employeeId and organisationId as primary key. I recommend to save the file in a cloud storage such as AWS S3 and have only the url to the file persisted in the database.

I recommend a document DB like MongoDB to persist the document info that enables us to query from any of the fields with ease.

Design

Document upload and verification should be an asynchronous system as the verification process is manual with an SLA of 1 to 4 hours. Latency is not a major concern and fetching the document status isn’t expected to be frequent for the same set of documents, hence there is no need to build a caching layer.

Follow up questions

Interviewer could optionally tweak the questions and ask the following questions

1. /createDocument uploads and initiates the verification process for a single document. What would you do if you were do bulk upload of documents or alternatively process pre existing set of many documents. The solution is batch processing. You can address it via one producer to many consumer processing via messaging queue or even background jobs that can be triggered both periodically and on demand manually.

2. Dive Deeper on Ticketing System if you were to build it on your own and not use an existing solution such as Jira.

3. How would you determine the ticket priority to ensure the SLA is met ? SLA could vary based on document type. Depending on balance time the ticket priority could be auto updated and alerted to the team.