Services
The client is a top global IT solutions provider serving Fortune 500 enterprises across financial services, healthcare, T&H, and insurance. The internal business unit-HROps, manages workforce transitions and background verification (BGV) processes for thousands of employees annually, ensuring compliance with client-specific and regulatory requirements.
BUSINESS CHALLENGE
HROps faced significant delays and manual overhead in verifying employee background data during redeployment. The BGV process involved extracting and comparing data from thousands of historical PDF reports stored in SharePoint, often resulting in 3-5 day delays per transition. These inefficiencies risked revenue leakage, non-compliance with client MSAs, and reputational damage, especially in regulated industries like BFSI and healthcare.
We implemented DoxPro, an AWS-native GenAI-based Intelligent Document Processing (IDP) platform. The solution automates document classification, key-value extraction, and semantic delta analysis using Amazon Bedrock (Anthropic Claude), Amazon Textract, and AWS Lambda.
Our solution met core automation goals, and we introduced pre-processing for metadata normalization and retraining prompt chains to help improve model precision and reduce manual intervention.
HR documents are stored in Amazon S3, triggering Lambda functions for processing
Claude on Bedrock performs deep semantic comparisons using its 100K-token context window.
AWS Step Functions manage fault-tolerant execution across services.
Delta results are stored in Amazon Aurora Serverless v2 for relational querying.
Amazon SNS and CloudWatch provide real-time alerts and operational insights.
Architecture
This modular, serverless architecture ensures high availability, scalability, and auditability across HR operations.
85% reduction in verification time
Average BGV processing time reduced from 3-5 days to under 24 hours.
3.5x increase in document throughput
Monthly processing capacity grew from 500 to over 1,800 employee records.
60% reduction in manual handling time
Complex HR cases are completed under 20 minutes, down from 45 minutes.
ARR impact
The solution generates an estimated $625K in AWS ARR, with Bedrock and Textract contributing over 39% of total usage.
Production-grade deployment
Fully operational across multiple business units with sustained usage of 1.5M tokens/day and 45K document comparisons/week