Ivan Yang Rodriguez Carranza
- 🎯 Problem Definition
- 🔬 Methodology
- 📝 Requirements
- 📋 Test Plan
- 🏛️ Architecture Diagram
- 🛠️ Technology Stack
- 🚀 Setup & Usage
- 🎬 Demo
- ✅ Hackathon Rule Compliance
- 📝 License
Organizations waste time and money on inefficient business processes because they lack the tools to identify where bottlenecks occur and how to fix them. Claimlytic uses intelligent claims analysis to automatically identify process inefficiencies and provide clear, actionable recommendations for improvement.
The project is built step-by-step through four main stages, as shown in the diagram below:
graph LR
A[📋 Requirements
+
📋Test Plan] --> B[🏗️ Architecture Design]
B --> C[⚙️ Implementation]
C --> D[🧪 Testing]
style A fill:#e1f5fe
style B fill:#f3e5f5
style C fill:#e8f5e8
style D fill:#fff3e0
The following table outlines all project requirements organized by category, with unique identifiers and priority levels to guide testing.
| Category | ID | Description | Testing Priority |
|---|---|---|---|
| 🔧 Functional | FR-001 | Accept claims data uploads in CSV, text, and Excel formats for analysis (accepts folder path to process all documents inside) | Low |
| 🔧 Functional | FR-002 | Accept PDF documents for business process documentation (accepts folder path to process all documents inside) | Low |
| 🔧 Functional | FR-003 | Identify process inefficiencies correctly and suggest clear, actionable recommendations for process optimization | Low |
| 🔧 Functional | FR-004 | Show explanations citing claims and business processes | Low |
| 🔧 Functional | FR-005 | Create "As Is" process flow diagrams showing current inefficient business processes | Low |
| 🔧 Functional | FR-006 | Create "To Be" process flow diagrams with optimized business processes | Low |
| 🔧 Functional | FR-007 | Enable users to ask questions through chat and clarify details to update the analysis | Low |
| ⚡ Performance | NFR-001 | Correctly identify 50% of the process inefficiencies | High |
| 🛠️ Technical | TR-001 | Use IBM Granite models for LLM inference | Low |
The testing strategy includes the following:
- Evaluations: Evaluate the main features (claims analysis, diagram generation, and AI suggestions) through real-world scenarios to assess practical effectiveness, ensuring the system correctly identifies 50% of process inefficiencies.
| Test ID | Test Case | Expected Outcome | Requirement ID |
|---|---|---|---|
| EV-001 | Process Inefficiency Detection Accuracy | Using a validation set of 4 known process improvement cases, system achieves ≥50% accuracy in identifying actual inefficiencies | NFR-001 |
Note: A more comprehensive evaluation that considers different industries and a larger number of test cases is required.
graph TB
G["📋 Claims"] --> A
I["📄 Organization documentation"] --> A
A["💬 Chat UI"] --> B
C --> J
J --> C
C --> A
J --> E
subgraph E["🔧 Provider"]
F["🌐 IBM watsonx API"]
H["💻 Local Granite LLM (Alternative)"]
end
subgraph B["⚙️ Service Layer"]
C["🔍 Analysis Service"]
J["🤖 LLM Service"]
end
style A fill:#fce4ec
style B fill:#e8f5e8
style C fill:#fff3e0
style E fill:#e1f5fe
style F fill:#f3e5f5
style G fill:#e1f5fe
style H fill:#f3e5f5
style I fill:#e1f5fe
style J fill:#f3e5f5
- AI/ML: IBM Granite
ibm/granite-3-3-8b-instruct(via IBM watsonx.ai) - Backend: Python 3.12+
- Document Processing: PyMuPDF
- Frontend: Streamlit
Note: While Claimlytic is configured to use IBM Granite models through watsonx.ai, it can be extended to use local IBM Granite models using the
LocalGraniteProvider.
- Python 3.12+
- IBM Watsonx AI account (for cloud-hosted IBM Granite models)
- UV package manager (optional, for UV installation method)
# Clone the repository
git clone https://github.com/rodcar/claimlytic.git
cd claimlytic
# Create virtual environment
python -m venv venv
# Activate virtual environment
# On macOS/Linux:
source venv/bin/activate
# On Windows:
# venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Run the application
python main.pyAlternative using UV Package Manager:
# Clone the repository
git clone https://github.com/rodcar/claimlytic.git
cd claimlytic
# Run the application (UV automatically handles dependencies and virtual environment)
uv run python main.pyFollow the steps on the Get Started page. Use the files from the data folder.
Create and configure .venv (Based on .venv.example) before running the tests.
# Using UV
uv run pytest -vLink: https://youtu.be/4YFdC4Xo6CI
- ✅ All AI/LLM functionality is powered by IBM Granite models (
ibm/granite-3-3-8b-instruct) via watsonx.ai. - ✅ No third-party LLMs, APIs, or proprietary services are used.
- ✅ All tooling is open source and self-hosted.
- ✅ Open-source.
Claimlytic is licensed under the Apache License 2.0.
© 2025 Ivan Yang Rodriguez Carranza.