ML based Automation for Work Permits and Fulfilment

About Tata Projects
Tata Projects is one of the fastest growing and most admired industrial infrastructure companies that offers procurement management services globally. Tata Projects also provides ready-to-deploy solutions for refineries, roads, bridges, integrated rail & metro systems, commercial building & airports, and power generation, transmission & distribution systems, chemical process plants, water & waste management, and mining & metal purification systems.

Customer Challenge
Tata Projects has over 20K staff allocated to customer locations. Tata Projects Limited has multiple contracts that are executed for their customers for which Work Permits are created. The Work Permits are created based on a Service request which is generated in the system.

To process the payments and track personnel working and track billing, these Work Permits are generated by the on-site supervisors who track the attendance of the personnel which includes their Gate Pass information, In and Out time.
Before starting this project, the central CSD team receive these Work Permits daily which include load of information that needs to be extracted manually and track the progress of a Service request. This is a time-consuming process and typically takes about 10-15 mins to process one Work Permit as most of the information needed for tracking this in the system (Internal CRM), required various mandatory information such as Skill Level, Shift Type etc. which are tracked from different systems.

The CSD team processes over 8000 to 9000 Work Permits per month which involves high number of resources and majority of this process can be automated by inducing a Machine learning model that can extract the prerequisite information and then APIs can fetch the relevant information for updating Service Requests.

Why AWS
Amazon Web Services offers a superior Platform that provides security and scale support through a broad architecture for multiple machine learning frameworks. With AWS services like Amazon Sagemaker a fully managed solution that offers cloud-based machine learning models to develop, train and deploy.

Why cloudmantra
cloudmantra has a proven record in developing several highly scalable and efficient Machine Learning models for various use cases across Manufacturing, Academia, BFSI, Media industries. Tata Projects requirements needed quick response time for implementing the machine learning use cases presented with high accuracy and required high technical expertise.

Solution
cloudmantra team has expertise to create ML based applications utilizing various services, libraries, and algorithms to design a fully automated solution.

Tata Projects generates about 8000 to 9000 Work Permits a month. These Work Permits are in form of PDF documents that are generated by the customers and handed over to the Supervisors on site. All these Work Permits are generated by an automated system and therefore are always available as soft copies too. Work Permits that were handwritten were discarded from this ML platform.
Each Work Permit format is standard and typically is two pages. Sometimes, if the number of contract staff on the Work Permits are higher, the pages may go into three pages.

With this available information, cloudmantra reviewed the Work Permit formats and the concerned audit team was interviewed for the parameters that they typically extract. Each information was noted as Mandatory or optional category for extraction from the form (Work Permits). The data coming from the extracted formats was also noted for variance.

A Sample of 200 Work Permits were used for extracting the information using Amazon Textract. The available and extracted information was then ingested to a Regular Expression function which segregated the required fields and values. Once this information was available, it was packaged as a json and updated to an Amazon S3 Bucket. These json files were then sent to the Tata Projects team to run tests for ingesting this data via their APIs.

The Tata Projects team and cloudmantra team worked to optimize the json format, include various information needed that could perfectly match what the human would update the Service requests in the CRM. Once the algorithm was able to generate the correct json, the task was completed.

The model was able to provide 100% accuracy of the extraction and generate an appropriate json object for the project.

Benefits
-Enabled focus on development not IT management thereby saving time
-Offered flexibility and agility to support deep analytics models thereby providing higher accuracy in results
-Ability to simultaneously train and deploy machine learning models under one platform thereby reducing management overheads and integrating with associated applications
-Built & trained models from collections data in weeks, not months thereby providing quick wins
-Increased efficiency by automating manual tasks thereby freeing human effort for audits from several hours to under 30 mins

AWS Services Used
-AWS Lambda
-Amazon Textract
-Amazon S3
-CloudWatch
-Amazon SageMaker Notebooks
-IAM