Trustflow

• 

  landing page

• 

  dashboard

Inspiration

Businesses lose countless hours manually verifying identities and matching payments to invoices. We wanted to build an agent that handles this end to end with humans only stepping in for edge cases.

What it does

Trustflow uses Qwen vision and reasoning models to automate client onboarding and payment reconciliation. Qwen-VL reads uploaded IDs and extracts data instantly while Qwen-Max verifies this data and scores its confidence to auto approve the client or escalate to a human. The agent also monitors incoming payments and matches them to invoices.

How we built it

We built the frontend with React Vite and Tailwind CSS. The backend runs on Node.js and connects to the Qwen Cloud API for all AI operations. We deployed the infrastructure on Alibaba Cloud using ECS for hosting OSS for document storage and RDS for our database.

Challenges we ran into

Building a reliable reasoning layer that knows exactly when to escalate to a human required careful prompting and validation logic.

Accomplishments that we're proud of

We are incredibly proud of successfully building a fully automated pipeline that handles both document extraction and decision making. Integrating the vision capabilities of Qwen-VL with the reasoning power of Qwen-Max to create a reliable system with a seamless human fallback is our biggest achievement.

What we learned

We learned a lot about prompting reasoning models for strict business logic and confidence scoring. We also gained valuable hands on experience deploying and scaling a full stack AI application on Alibaba Cloud infrastructure.

What's next for Trustflow

We plan to expand the system to support a wider variety of international documents and languages. We also want to integrate directly with popular accounting software to make the payment reconciliation process even smoother.

Built With

• alibaba-cloud
• node.js
• qwen-max
• qwen-vl
• react
• tailwind-css
• typescript