Risk Management
Base39 needed to replace slow, subjective credit analysis workflows that could take up to 10 days and required specialist intervention to adapt machine learning models to new data sources.
Reported impact
Base39 deployed its AI-driven credit analysis platform in two weeks, cut loan-analysis costs by 96%, lowered infrastructure spend by 84%, and moved toward analyzing up to 10 loan applications per minute.
Base39 rebuilt its credit analysis workflow on MongoDB Atlas running on Amazon Web Services and used MongoDB Atlas Vector Search as the feature and retrieval layer for agentic AI. Data from CRM systems, transactional records, and public databases is aggregated into Atlas, where flexible schemas let the team mix structured financial attributes with unstructured context. LangGraph orchestrates agents that call predictive machine learning models plus Amazon Bedrock-hosted Claude 3.7 Sonnet and Claude 3.5 Haiku models to recommend how data should be cleaned, transformed, and evaluated for loan risk. Atlas Vector Search enriches those LLM interactions with domain-specific context, supports real-time feature retrieval, and enables Chain-of-Thought reasoning over each application. Because the stack remained familiar to developers, Base39 shipped quickly while eliminating manual spreadsheet-driven analysis.
CRM records, transactional histories, and public-database signals are aggregated into MongoDB Atlas, embedded and indexed with MongoDB Atlas Vector Search, then passed through LangGraph-orchestrated Bedrock model calls so agents can score loans, run what-if analysis, and surface risk recommendations to analysts.
Base39 cut loan analysis costs by 96% and reduced infrastructure expenses by 84%.
“MongoDB Atlas serves as the backbone of Base39’s feature store, powering both structured and unstructured data analysis with MongoDB Atlas Vector Search.”
“MongoDB underpins the data layer, providing flexible schema support, vector search for LLM context, and a managed deployment model that aligns with Base39’s developer-first philosophy.”
“The strategic partnership with MongoDB and AWS enabled Base39 to deploy a robust AI-driven solution in two weeks.”
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
MongoDB Sample Jun 24, 2026 ani.tech ani.tech needed an AI-native data layer that could support autonomous financial-workflow agents after MySQL and rigid SaaS-era schemas became a bottleneck for experimentation and deployment. Its automated agents absorbed 75% of an end-to-end financial-advice workflow for pilot customer Octopus Money. MongoDB Atlas MongoDB Sample Archived · Feb 10, 2025 XTransfer XTransfer's long transaction chains, fragmented global data, and growing compliance obligations required a more flexible database that could scale cross-border payments while supporting real-time risk controls and… over 200 countries and regions ApsaraDB for MongoDB 
OpenAI Sample Oct 16, 2026 Benefit Street Partners Benefit Street Partners needed to answer investment-committee follow-up questions faster, compare lender marks across shared loans, and review 300-page credit agreements for clauses that could shift value away from… 10–12 hours total. BDC mark tracker · Codex · Credit agreement analyzer 