Community

Blog Events Webinars Tutorials Forum
  1. Blog
  2. Events
  3. Webinars
  4. Tutorials
  5. Forum
Create Account
×
Community Blog Building FinTrack: An AI-Powered Tax Assistant for Freelancers Using Alibaba Cloud

Building FinTrack: An AI-Powered Tax Assistant for Freelancers Using Alibaba Cloud

This article explains how to build FinTrack, an AI-powered tax assistant for freelancers, using Alibaba Cloud's generative AI platform and its Qwen large language models.

1

Freelancers are among the fastest-growing segments of the global workforce. But while their ways of working have evolved, the tools supporting them - particularly for financial management - have not. Many platforms overcomplicate tax filing, while others offer limited automation for tracking expenses across borders and currencies.

At the Malaysia AI Hackathon 2025, one team set out to address this gap.

Meet FinTrack: a fully cloud-native, AI-powered expense and tax assistant built to simplify financial workflows for freelancers, from receipt upload to tax deduction insights - using Alibaba Cloud’s full-stack AI and data tools.

The Problem: Freelancers Need Smarter Financial Tools

Today’s freelancers often:

  • Handle receipts and invoices in multiple currencies
  • Miss out on eligible deductions due to manual processes
  • Lack access to intuitive, AI-assisted financial insights

The ideal solution would:

  • Automatically extract and categorize expenses from receipts
  • Offer real-time income summaries and deduction suggestions
  • Visualize tax progress in a single, user-friendly dashboard

The Architecture

Team 214 adopted a serverless-first architecture, leveraging Alibaba Cloud’s AI, database, and analytics products to rapidly build and deploy FinTrack during the hackathon.

1. OCR + Auto Categorization

From image upload to structured financial data

2

  • Users upload a receipt through the frontend
  • Images are stored in Object Storage Service (OSS)

  • Qwen-VL-Max, a multimodal LLM from Model Studio, is prompted with predefined tax-relief categories to extract:

    • Date, vendor, currency, type, amount, and description
  • Categorized results are stored in ApsaraDB RDS for MySQL

2. Tax Summary & Insights Engine

From raw transactions to tax-saving suggestions

3

Components:

  • Transactions are grouped and enriched with personal income context

  • Qwen-Plus is used to:

    • Summarize income and expenses
    • Generate tax deduction suggestions

Results are visualized using Quick BI, delivering:

  • Real-time tax summaries
  • Category-level progress indicators
  • Deduction breakdowns for filing

Working Within the Alibaba Cloud Ecosystem

As the hackathon required participants to build their solutions using Alibaba Cloud, Team 214 approached the challenge as an opportunity to explore how the platform's components could support a complete AI application - from ingestion to insight.

Throughout development, a few elements proved especially helpful:

  • Model Studio with Qwen enabled the team to handle both image-based OCR and natural language processing tasks using large language models in a single environment.
  • OSS (Object Storage Service) simplified image handling, allowing receipt uploads to be securely stored and accessed by their AI pipeline.
  • ApsaraDB RDS (MySQL) served as a solid foundation for organizing structured transaction data with minimal configuration.
  • Quick BI helped transform backend data into a usable dashboard interface quickly - a major advantage in a time-constrained build.

4

While the team initially explored these tools out of necessity, they found the platform’s consistency and built-in integrations allowed them to move faster and spend more time refining the user experience.

Engineering Highlights

  • Prompt-based categorization: Instead of hardcoding tax rules, we prompted Qwen with a dynamic system of categories and thresholds. This keeps the logic flexible across countries.
  • Semantic receipt analysis: Qwen-VL-Max can read poorly formatted receipts, interpret amounts in multiple currencies, and identify non-obvious tax-deductible items.
  • BI without frontend: Quick BI removed the need to build a full analytics UI. We simply connected the RDS and configured user-level filters.

FinTrack is a strong example of what’s possible when real-world problems meet innovation.

Using Alibaba Cloud’s products, Team 214 was able to go from idea to implementation by combining AI, automation, and analytics into a streamlined user experience for freelancers in Malaysia.

Build with Purpose, Powered by AI

FinTrack started with a simple question: How can we make life easier for freelancers?

The answer came to life during the hackathon - with creativity, collaboration, and the right tools.

If you have an idea that can make a difference - whether it’s in finance, health, sustainability, or education - the Alibaba Cloud Global AI Hackathon is where it can take shape.

  • Use cutting-edge tools like Qwen and Model Studio
  • Collaborate with a global network of innovators
  • Bring your project to life in an environment built for builders

Your idea could be next, Join us and build something meaningful!

AI Artificial Intelligence Developers cloud native Generative AI GenAI LLMs Qwen AI Hackathon Model Studio
0 1 0
Share on

Read previous post:

GoGuard: Giving Peace of Mind to Late-Night Riders Through AI-Powered Safety

Read next post:

Empowering SMEs with AI: How FundSight AI is Transforming Financial Clarity on Alibaba Cloud

Alibaba Cloud Project Hub

134 posts | 26 followers

You may also like

Comments

Alibaba Cloud Project Hub

134 posts | 26 followers

Related Products

More Posts by Alibaba Cloud Project Hub

See All