The following is a take-home project for Soraban engineering candidates.

• Repository — give read access to @seanmcoleman with a README that covers setup
• Video Walkthrough — Loom (or other video) showcasing the functionality you built and explains how the code works

Build a minimal yet scalable bookkeeping system with the following features:

1. Record & Import Transactions – Users can manually add transactions or import a CSV.
2. Bulk Actions & Automated Categorization – Users can categorize multiple transactions at once, and automatically assign category (AI-based or rule based).
3. Anomaly Detection – Identify and flag unusual/suspicious transactions (e.g., large amounts, duplicates, missing metadata).
4. Scalability & Performance Optimization – Efficiently handle large data sets (e.g., 1m+ transactions).
5. User-friendly Review System – A dashboard that highlights transactions needing review.

• Backend: Ruby on Rails (preferred), Node.js, Django, or similar.
• Frontend: React (preferred) or Vue.js.
• Database: PostgreSQL (preferred) or MySQL.

• Users can manually add transactions (date, description, amount, category).
• Users can import a CSV file containing transactions.
• CSV parsing should handle edge cases (missing fields, malformed data, duplicates).

• Users can select multiple transactions and apply bulk categorization.
• Users can create rules like:
  • “If the description contains ‘Amazon’, categorize as ‘Shopping’.”
  • “If amount > $1000, flag as ‘High Value’.”
• Rules should apply automatically when new transactions are added.

• Identify transactions that are:
  • Unusual in amount compared to past user behavior.
  • Potential duplicates (same amount, date, with same descriptions).
  • Incomplete/missing metadata (e.g., description missing).
• Flag these anomalies and display them on the Review Dashboard (Step 5)

• Your system should handle 1m+ transactions efficiently.
• Consider indexing, caching, or batch processing for performance.

• A simple dashboard that highlights:
  • Uncategorized transactions needing user review.
  • Flagged anomalies requiring manual verification.
• Users should be able to approve, edit, or delete flagged transactions.

1. Basic API for Transactions – Expose a REST API for CRUD operations.
2. Real-time Anomaly Detection – Use WebSockets or polling for updates.
3. Graph-based Spending Summary – Show user spending trends.

✅ Code Quality & Architecture – Clean, modular, and scalable.

✅ Performance & Efficiency – Handles large datasets without slowdowns.

✅ Complex Logic Implementation – Anomaly detection & rules engine.

✅ Good UX for Complex Actions – Well-designed transaction review.

✅ AI Resistance – Requires thoughtful business logic, rule handling, and anomaly detection, which AI struggles to generate effectively.

✅ Problem-Solving Skills – Ability to balance features, scalability, and performance.

• Kick.co
• Quickbooks Online
• Xero