SemantiPay

• 

  SemantiPay Hero Section

• 

  Payroll Dashboard

• 

  Processed Payroll Results

• 

  Architecture & Tech Implementation

Inspiration

Payroll compliance needs hours of developer work every time rules change. SemantiPay cuts that effort using AI.

What it does

• Automates payroll processing

• Stays compliant with global rules

• Skips irrelevant compliance rules and understands meaning, not just keywords

How I built it

• TiDB as both SQL database and vector store

• Dify for workflow automation

• Leveraged AI models/services: gpt-oss-120b (via Groq), Jina AI embeddings

Workflow

• The system fetches employee payroll data and generates compliance queries with embeddings. Using semantic search, it retrieves the relevant compliance rules and processes payroll using AI. If a violation is detected, the system notifies HR by sending an email & a Slack message.

Challenges I ran into

• Dify couldn’t query TiDB vectors directly from tables. (Workaround: Developed an Agentic Payroll Middleware to mainly integrate Dify & TiDB)

What I learned

• Hands-on with new tech (TiDB, Dify, Jina AI, Groq)

• Insights into payroll compliance

• Planning in solo development

What’s next for SemantiPay

• Document upload & parsing feature using AI for new compliance rules

• Compliance chatbot

• Personalized HR dashboards (Multi-tenancy)

Built With

• dify
• express.js
• gpt-oss-120b
• groq
• javascript
• jina
• next.js
• redis
• tidb
• typescript