Project #2
ChatFill PDF: AI-Powered Intelligent Document Filling SaaS
Role: Founder & Lead Developer
Tech Stack: React, n8n (Workflow Orchestration), OCR (PDF Processing), LLM.
The Challenge:
- Traditional PDF form filling is time-consuming and error-prone, with users often abandoning the process due to tedious workflows.
- Need to solve the precise mapping problem between unstructured natural language (User Input) and structured form fields (PDF Fields).
The Solution:
- Built an AI Agent Workflow: first use OCR to parse PDF structure, then leverage LLM semantic understanding to convert user conversations into JSON data, finally auto-fill the PDF.
- Implemented Schema Matching algorithm to handle complex address formats, date formats, and multi-level field mappings.
Key Achievement:
Reduced form filling time from 20 minutes to 5 minutes (75% efficiency improvement).
Completed MVP development and entered Go-to-Market phase.
Project #3
AI Voice Sales Agent: Fully Automated Real Estate Voice Sales System
Role: AI System Integrator
Tech Stack: Vapi (Voice AI), Custom CRM, Vector Database (Sales Training Data).
The Solution:
- Built an End-to-End voice sales loop: CRM trigger → AI calling → intent recognition → auto-scheduling/CRM update.
- Advanced negotiation capability: vectorized NEPQ sales training to enable AI to handle customer objections.
- Human-in-the-loop design: AI handles front-end screening and warm-up, seamlessly transferring high-intent customers to human experts, significantly improving conversion rates.
Project #4
Central Bank Simulator: Multiplayer Online Economic Game
Role: Full Stack Developer & Game Designer
Tech Stack: React, Node.js, WebSocket (Socket.io for Real-time Sync).
The Challenge:
Need to simulate real-world economic interconnections and handle high-concurrency state synchronization for 6 simultaneous online players.
The Solution:
- Implemented WebSocket protocol to ensure all players' decisions (such as raising/lowering interest rates) are broadcast in real-time and update game state.
- Designed numerical models based on real economics (such as Taylor Rule) to simulate the interconnections between inflation, GDP, and policy tools.
Key Achievement:
Successfully transformed dry monetary banking theory into a highly interactive multiplayer game, demonstrating the ability to handle real-time data streams.
Project #5
AI Insurance Claims Calculator: Insurance Claims Estimation Tool
Role: Frontend Developer & Logic Architect
Tech Stack: React, LLM (Semantic Analysis).
The Solution:
01
Natural Language Understanding
Utilize LLM for natural language understanding to automatically match user-described accident scenarios to corresponding insurance clauses.
02
Insurance Logic Engine
Implemented a complex insurance logic engine: handling 'no duplicate reimbursement', 'deductible calculation', and 'multiple policy stacking' and other practical constraints.
03
Clause Mathematization
Transform unstructured policy clauses into computable mathematical formulas.
