Building a SaaS from Mortgage Challenges: The Lektik Way

Client Overview 

A mortgage and loan services provider specializing in origination, underwriting, and approval approached Lektik with a significant bottleneck: slow, manual document handling that delayed processing, increased operational costs, and posed compliance risks. What began as a need for automation evolved into the creation of a scalable B2B SaaS company — with Lektik serving as both technology partner and venture builder. 

Challenges 

• Manual Document Verification 
  Reviewing diverse documents (bank statements, tax returns, income proofs) was time-consuming and error prone.

• Inconsistent Data Extraction 
  Loan officers manually pulled data from unstructured files like scanned images and PDFs, leading to frequent errors. 

• High Operational Costs 
  Manual workflows required a larger team, driving up overhead and limiting efficiency. 

• Compliance & Regulatory Risk 
  Human-led processes risked missing regulatory checks, leaving the business vulnerable. 

• Scalability Constraints 
  The manual approach couldn’t scale with increasing loan application volumes. 

Lektik’s Role as Venture Studio 

Innovation & Product Development 

• Developed a modular backend using Java Spring Boot and Python FastAPI for robust, microservice-based architecture. 

• Implemented intelligent document processing using: 

• Paddle OCR, Sparrow OCR, and Tesseract for multi-layered OCR precision. 

• DeepSeek LLM for context-aware classification, validation, and key data extraction. 

• Built MISMO XML parsers and validation layers to ensure structured, compliant data capture. 

• Enabled simultaneous document processing and improved turnaround using multithreaded pipelines. 

Technology Company Launch 

• Incubated and spun off a new SaaS product from the client’s internal toolset. 

• Created a public-facing application and website for external adoption. 

• Supported regulatory setup and product packaging for compliance in the mortgage tech space. 

Go-to-Market Enablement 

• Facilitated cloud infrastructure and compliance partnerships (AWS stack: Cognito, S3, CloudWatch, ECS, ECR). 

• Integrated with industry tech partners and supported affiliate onboarding. 

• Co-developed sales materials and onboarding processes for early clients. 

Business Impact 

Metric	Result
Document Processing Time	↓ 70%
Manual Data Entry	↓ 85%
Data Extraction Accuracy	↑ 93%
Data Entry Errors	↓ 89%
Loan Capacity (same staff)	↑ 60%
Regulatory Compliance	↑ 100%
Loan Volume (no new hires)	↑ 40%
Underwriting Costs	↓ 30%
Approval Speed	↑ 50%

(Note: internal projections based on early user behavior and estimates from initial testing) 

Technologies Used 

• Backend: Java Spring Boot, Python FastAPI 

• Frontend: React.js 

• Cloud & Infrastructure: AWS (Cognito, S3, CloudWatch, ECS, ECR) 

• Messaging Architecture: RabbitMQ 

• Document Intelligence: PaddleOCR, SparrowOCR, Tesseract, DeepSeek LLM 

The Road Ahead 

Work is ongoing to enhance AI models, expand document coverage, and enable end-to-end generation of compliant MISMO XML files.  

With Lektik continuing to support product evolution, the platform is well-positioned to become a category-defining solution in mortgage tech. 

---