Build Autonomous Home Improvement Loan Underwriting Pipeline with Mastra AI
Evaluating $3 billion in home improvement loans requires asynchronous, fast TypeScript workflows capable of processing borrower income, contractor bids, and property tax records concurrently. Developers will construct an autonomous underwriting pipeline using Mastra AI. The pipeline parses incoming loan packages, validates contractor credentials against state licensing databases, verifies debt-to-income (DTI) compliance, and outputs instant approval, decline, or conditional manual review decisions.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an asynchronous home improvement loan underwriting workflow agent using Mastra AI in TypeScript.
How work is evaluated
Evaluates underwriting decision accuracy and enforcement of DTI threshold limits.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
DTI Limit Enforcement
Verifies that DTI > 43% results in DENIED status
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Decision Accuracy
Underwriting disposition accuracy against rules • target: 1 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Configure Mastra AI TypeScript workflow steps for asynchronous loan evaluation
Build dynamic contractor verification tools to check active trade licenses
Enforce DTI and CLTV underwriting limits using structured step conditionals
Output standardized JSON loan disposition decisions
Reference links and supporting material
Dataset of 50 consumer home improvement loan applications with income verification and contractor details.
Challenge setup
Evaluates underwriting decision accuracy and enforcement of DTI threshold limits.
Challenge input
JSON with monthly_income, total_monthly_debt, requested_loan_amount, contractor_licensed
Mastra AI
Required framework for TypeScript workflow agent.
Evaluated output
JSON with decision ('APPROVED', 'DENIED', 'MANUAL_REVIEW'), calculated_dti, rejection_reasons
- Verifies that DTI > 43% results in DENIED status
- Underwriting disposition accuracy against rules • target: 1 • range: 0-1
- Benchmark: UnderwritingWorkflowBench
- Decision Accuracy target: 1
- 1 public reference case
- JavaScript execution harness
- javascript sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Mastra AI · Required
- Cartesia · Optional
- Portkey · Optional
- Portkey · Optional
- Mastra AI · Required
Evaluation contract
- DTI Limit Enforcement · Weight 1
- Decision Accuracy · Weight 1
Metrics
- Decision Accuracy · Target 1
Benchmarks
- UnderwritingWorkflowBench
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON with monthly_income, total_monthly_debt, requested_loan_amount, contractor_licensed
Output: JSON with decision ('APPROVED', 'DENIED', 'MANUAL_REVIEW'), calculated_dti, rejection_reasons
Execution harness: JavaScript
Evaluation tools
- Mastra AI · Agent Framework
Recipe freeze marker
No recipe freeze time is recorded.