Enroly

ABOUT:

Enroly provides the first CAS, visa and arrival automation platform. It makes post-acceptance administration simple and secure, allowing universities to easily implement the solutions into their existing systems. Not only does it help improve enrolment systems, but it also automatically detects risk factors to help higher education institutions to prevent mistakes that would otherwise become deferrals, withdrawals and refusals. Enroly's range of products are used by universities, students, agents, and regional offices in more than 100 countries around the world.

INDUSTRY:

Education

COUNTRY:

UK

Partnership with STX Next:

February 2020 - April 2020 (2 months)

Project:

Improving Enroly CAS SHIELD with Machine Learning

Enroly CAS SHIELD, one of the company's three core services, helps university staff manage applications from prospective students. By offering functionalities such as smart prioritization of large applicant volumes, student interview booking system, and automated reading and assessing of documents, the company helps improve enrolment numbers. The platform also automatically detects risk factors and assigns enrolment confidence scores to applications. STX Next collaborated with Enroly to increase efficiency and effectiveness with machine learning solutions, including by automating document processing and information extraction.

STX NEXT DELIVERED

Team Extension

Team Extension

Machine Learning

Machine Learning

Data Engineering

Data Engineering

Backend Development

Backend Development

Continuous Integration

Continuous 
Integration

Automated Testing

Automated
Testing

Manual Testing

Manual
Testing

DevOps

DevOps

Integration Testing

Integration Testing

TECH STACK

Python logo

Python

AWS logo

AWS

Docker logo

Docker

AsyncIO logo

AsyncIO

Amazon DynamoDB logo

Amazon DynamoDB

Boto3 logo

Boto3

AWS Lambda logo

AWS Lambda

Tesseract OCR logo

Tesseract OCR

...and Amazon S3, Slack, Jira, CloudWatch, GitHub, SciPy, Numpy, Amazon Comprehend, AWS CodeBuild, Amazon Textract

CHALLENGE:

Since Enroly's platform hosts large volumes of documents, the company was looking for ways to make their processing more efficient, free up their clients' employee time, as well as speed up the verification of documents and the application for the CAS statement.

The first challenge, therefore, was to identify the specific areas that would benefit from the introduction of machine learning, and create a workable project scope. 

Once the scope was established, Enroly needed the right tools to help it process and extract the relevant information from the huge range of documents it receives from all over the world (for instance birth certificates, bank statements, and handwritten notes) in varying quality and formats (for instance scans and photos). 

 

SOLUTION:

STX Next held consultations with Enroly to find areas that could be automated with ML solutions as well as explore potential features and prioritize them based on their business value. 

Following the consultations, we created a project scope which served as a roadmap for future development of AI-based features.

The scope consisted of the application of Optical Character Recognition techniques to automatically process and verify documents submitted to the platform by students or agencies acting on their behalf. The documents were analyzed and relevant data was extracted from them. Natural Language Processing techniques were used to implement the fuzzy matching method, which helped ensure the required information has been provided.

TESTIMONIALS

Author

Our team was tasked with developing features for automated extraction and processing of data from scans of various documents. The development process included implementing, testing, and comparing many ML methods to deliver the best possible end solution. I will remember our cooperation with the client as an enjoyable experience. We could always rely on great communication and fast decision making.

— Michał Wiśniewski, Machine Learning Engineer @ STX Next

Request our Introduction Pack

with 40+ pages of projects and full tech stack info: