MSME Risk Intelligence Agent

AI agent that turns MSME borrower documents into instant risk scores, early warning signals, and automated review workflows for faster, auditable lending decisions.

Comment

🛡️ Shield RiskOps: The SDLC Compliance & Planning Agent The Problem

In FinTech and lending, the biggest bottleneck in the software development lifecycle is often not coding, but compliance and planning. Before teams can deploy lending-related features or move forward with business workflows, they often depend on manual review of large, unstructured financial and regulatory documents such as PDFs, Word files, and bank statements. This slows decisions, creates friction, and delays execution.

Our Agentic AI Solution

Shield RiskOps is a proactive, autonomous GitLab Duo custom agent built to reduce this compliance bottleneck. Instead of acting like a reactive chatbot, it functions as a digital compliance auditor embedded directly into the workflow.

How it Works Parse Agent Uses Google Gemini 2.5 Flash to read unstructured financial inputs such as PDFs and CSVs, then converts them into a strict structured JSON schema. Risk Scorer Agent Computes financial trends and deterministically identifies critical risk patterns such as sharp revenue decline, missed EMI events, or bounced payments. Compliance Agent Generates detailed, audit-ready markdown reports summarizing the borrower’s risk profile and findings. Action Agent Uses external tools to take real operational action by: creating a live GitLab review issue with a pre-filled checklist sending an SMTP email alert to the branch manager with the compliance audit attached Why it Matters

By converting raw borrower data into structured risk analysis and immediately pushing the result into GitLab and email workflows, Shield RiskOps helps teams move faster on compliance decisions without waiting for slow manual reviews.

GitLab Duo Custom Flow

To align with the GitLab Duo Agent Platform, the agent flow and tool configuration are defined in:

.gitlab/agents/shield-riskops/flow.yml

This connects the LLM’s workflow to GitLab tool actions such as issue creation and operational escalation.

Final Impact

Shield RiskOps demonstrates that AI can do more than answer prompts. It can act as an autonomous digital teammate that reduces SDLC compliance friction, accelerates planning, and takes concrete operational action.

Built With

  • ci/cd
  • gemini-2.5-flash
  • gitlab
  • gitlab-api
  • google-gemini
  • pypdf
  • python
  • python-docx
  • requests
  • smtp
  • streamlit
  • yagmail
Share this project:

Updates