Exclusive Content:

Haiper steps out of stealth mode, secures $13.8 million seed funding for video-generative AI

Haiper Emerges from Stealth Mode with $13.8 Million Seed...

Running Your ML Notebook on Databricks: A Step-by-Step Guide

A Step-by-Step Guide to Hosting Machine Learning Notebooks in...

“Revealing Weak Infosec Practices that Open the Door for Cyber Criminals in Your Organization” • The Register

Warning: Stolen ChatGPT Credentials a Hot Commodity on the...

Sun Finance Enhances ID Extraction and Fraud Detection Using Generative AI on AWS

Transforming Identity Verification and Fraud Detection at Sun Finance with AWS

A Collaborative Innovation Journey with the AWS Generative AI Innovation Center

Overcoming Challenges in Identity Document Processing

Building a Robust AI-Powered Solution: Key Components and Architecture

Enhancing ID Extraction and Fraud Detection: Detailed Methodologies

Measurable Results: Improved Accuracy, Cost Efficiency, and Operational Impact

Future Directions: Scaling and Enhancing Fraud Detection Capabilities

Key Takeaways and Lessons Learned from the Implementation

Next Steps: Expanding and Improving the AI-Driven Solutions

Conclusion: Delivering Fast and Efficient Service to Microloan Applicants

Revolutionizing Identity Verification in Fintech: How Sun Finance Transformed Document Processing

This post was co-authored with Krišjānis Kočāns, Kaspars Magaznieks, and Sergei Kiriasov from Sun Finance Group.

In the fast-paced world of fintech, processing identity documents effectively is a cornerstone of operational efficiency. For organizations handling vast numbers of loan applications, account openings, and compliance checks, traditional optical character recognition (OCR) often presents significant challenges. While OCR is a useful tool, it frequently leads to extraction errors, thereby forcing many applications back into the realm of manual review—a labor-intensive, costly process that hinders scalability and efficiency.

Enter Sun Finance, a Latvian fintech founded in 2017 that has established itself as a technology-driven leader in online lending across nine countries. Faced with approximately 80,000 monthly applications for microloans—60% of which required manual review—Sun Finance partnered with the AWS Generative AI Innovation Center to redesign their identity verification pipeline, setting a new standard for efficiency. This collaboration yielded remarkable results, including a drastic reduction in processing costs and an impressive increase in accuracy.

The Project Timeline: From Kickoff to Launch

The project unfolded over 107 business days, culminating in a production launch in January 2026. Key milestones included:

  • Proof-of-Concept: From August 26 to October 9, 2025, the solution was fleshed out over 32 days.
  • Technical Handover: A 26-day process followed on November 14, 2025.
  • Production Launch: Following a 14-day holiday freeze, the solution went live on January 22, 2026.

The Identity Verification Challenge

Initially, Sun Finance relied on its first ID verification automation built using Amazon Rekognition and Amazon Textract. However, as the company expanded, the limitations of this system became evident. The local complexities of language and document formats contributed to a high error rate in extraction, causing a significant backlog in manual reviews.

Analysis revealed that 60% of manual interventions stemmed from mismatches due to OCR errors—a staggering demonstration of the insufficient capabilities of traditional OCR systems. Coupled with the need for robust fraud detection mechanisms, the manual workload became increasingly untenable.

Enter AI-Powered Solutions

The ambitious initiative with AWS led to two critical innovations: an advanced ID extraction system and a parallel fraud detection methodology. The innovative design utilized various AWS services, including:

  • Amazon Bedrock: For AI structuring
  • Amazon Textract: For OCR text extraction
  • Amazon Rekognition: As a fallback OCR system and for face detection
  • Amazon S3 Vectors: For serverless vector similarity searches

This synergy of technologies enabled Sun Finance to transform its document processing landscape significantly.

Achieving New Heights in Performance

By utilizing a specialized approach, Sun Finance experienced outstanding improvements:

  • Extraction Accuracy: Jumped from 79.7% to 90.8%
  • Cost Reduction: Achieved a staggering 91% decrease in per-document costs
  • Processing Time: Reduced from 20 hours to under 5 seconds per document

This innovative architecture combined specialized OCR capabilities with large language model (LLM) structuring, demonstrating a superior method over utilizing traditional systems alone.

Lessons Learned

The project yielded five essential insights:

  1. Integration of Technologies: Utilizing OCR and LLM together outperformed any single solution.
  2. Multi-Tiered Approaches: Employing multiple OCR tools created resilience against edge cases.
  3. Diverse Fraud Detection Methods: Both visual pattern analysis and background similarity checks are vital for comprehensive fraud detection.
  4. Incremental Complexity: Starting simple and progressively adding complexity based on analytics led to significant cost savings.
  5. Rapid Iteration with Serverless Architectures: Leveraging AWS’s serverless capabilities allowed for quick modifications and deployments without downtime.

Road Ahead: Expansion and New Opportunities

Looking to the future, Sun Finance plans to expand its operations further, refining its fraud detection capabilities and integrating additional signals. By harnessing the insights gained from this project, they aim for continued scalability, flexibility, and adoption across diverse markets.

In conclusion, the collaboration between Sun Finance and AWS has paved the way for a next-generation identity verification process, showcasing how innovative technologies can drastically improve speed, accuracy, and cost efficiency in fintech operations. This endeavor exemplifies how generative AI can transform foundational workflows, ensuring that applicants receive timely evaluations while simultaneously fortifying anti-fraud mechanisms.

Join the Conversation!

We encourage you to share your thoughts and experiences with document processing and fraud detection in the comments below. How could these technologies impact your organization?


For more information on generative AI in document processing and the solutions offered by AWS, visit the Amazon Bedrock product page or connect with the AWS Generative AI Innovation Center. Let’s reshape the future of fintech, one application at a time!

Latest

Request to Convert Business Unit into Educational Space Denied

Planning Application for Business Unit Conversion Denied by Bannau...

AWS Transform Now Automates BI Migration to Amazon Quick in Just Days

Accelerating Your Migration to Amazon Quick: Transforming Legacy BI...

Introducing DeepSeek V4: A Comparison with ChatGPT, Claude, and Gemini

The AI Arms Race: DeepSeek V4 Emerges as a...

Meta Acquires Robotics Startup Assured Robot Intelligence to Boost Humanoid Initiative Amidst $5 Trillion Market Competition

Meta Platforms Expands Humanoid Robotics Ambitions with Acquisition of...

Don't miss

Haiper steps out of stealth mode, secures $13.8 million seed funding for video-generative AI

Haiper Emerges from Stealth Mode with $13.8 Million Seed...

Running Your ML Notebook on Databricks: A Step-by-Step Guide

A Step-by-Step Guide to Hosting Machine Learning Notebooks in...

VOXI UK Launches First AI Chatbot to Support Customers

VOXI Launches AI Chatbot to Revolutionize Customer Services in...

Investing in digital infrastructure key to realizing generative AI’s potential for driving economic growth | articles

Challenges Hindering the Widescale Deployment of Generative AI: Legal,...

Snap Cuts 1,000 Jobs Amid AI Integration

Snap's Workforce Reduction: A Strategic Shift Towards AI-Driven Efficiency In this article, we explore Snap's recent layoff announcement, revealing the company's shift toward smaller, AI-augmented...

Reinforcement Fine-Tuning Using LLMs as Evaluators | Artificial Intelligence

Enhancing Large Language Models: The Power of Reinforcement Fine-Tuning with LLM-as-a-Judge Introduction to Reinforcement Fine-Tuning Benefits of RFT with LLM-as-a-Judge Key Steps for Implementing LLM-as-a-Judge Selecting the Right...

Allbirds Embraces AI: Exploring the NewBird Transformation

Allbirds' Bold Transformation: From Sustainable Sneakers to AI Compute Infrastructure From Wool to Wires The NewBird AI Announcement The Stock Reaction A Pattern Worth Watching What Allbirds Actually Built Conclusion:...