D-Cipher AI is an enterprise-grade analytical platform designed to provide transparent, rule-based intelligence for high-stakes decision-making. By integrating Neuro-Symbolic logic with asymmetric optimization, the platform transforms complex "black-box" datasets into human-readable, actionable insights.

Originally developed for Industrial Predictive Maintenance (Remaining Useful Life estimation), the system has evolved into a domain-agnostic engine capable of auditing diverse scenarios, from Aerospace engineering to HR Fairness monitoring.

graph TD
    User((User)) -->|Interacts| UI[Next.js Dashboard]
    UI -->|API Requests| API[FastAPI Backend]
    
    subgraph "Neuro-Symbolic ML Core"
        API -->|Trigger Extraction| NS[Neuro-Symbolic Engine]
        NS -->|XGBoost| FE[Feature Selection]
        NS -->|GMM| FU[Gaussian Fuzzification]
        NS -->|Interval Partitioning| RE[Rule Extraction]
        RE -->|Quantile Regression| OR[Optimization & Risk Profile]
    end
    
    OR -->|Interpretable Rules| API
    API -->|JSON Payload| UI
    
    subgraph "Advanced Analytics Suite"
        UI -->|Simulation| OL[Optimization Lab]
        UI -->|Bias Detection| AD[Ethical AI Audit]
        UI -->|Summarization| AA[AI Maintenance Analyst]
    end

The core engine utilizes a hybrid approach to bridge the gap between performance and interpretability:

• Feature Importance: Leverages XGBoost Gain to identify critical signal contributors.
• System 2 Partitioning: Implements recursive interval partitioning to extract logical antecedents in human-readable formats.
• Asymmetric Risk Profiling: Uses Quantile Regression (Pinball Loss) to allow users to shift the model's objective between aggressive performance and safety-critical conservatism.

The Optimization Lab provides a "Theoretical Limit" benchmarking tool. It bypasses interpretability constraints using deep decision trees and high-speed Ridge regression to calculate the maximum possible performance gain. This provides a clear comparison against the interpretable rule set, allowing users to quantify the "accuracy vs. explainability" trade-off.

The platform implements state-of-the-art evaluation metrics for fleet-wide datasets:

• Fleet-Wide Evaluation: Analysis is performed across 100% of the test set, ensuring that metrics like RMSE are representative of the entire population rather than single entities.
• Relative Scalability: All performance indicators are automatically scaled against the global maximum lifespan or target range, providing essential context for result interpretation.
• SOTA Verification: The engine consistently matches or outperforms industry-standard benchmarks (e.g., NASA C-MAPSS).

D-Cipher AI features a deterministic, high-performance UI inspired by modern engineering tools:

• Design System: A premium dark-mode interface with motion-stabilized components and Vercel/Linear aesthetics.
• Hydration-Stable Rendering: Optimized React mounting strategies to eliminate flickering and hydration mismatches.
• High-Fidelity Visualizations: Real-time trajectory tracking for multiple representative entities (Top 3), allowing for immediate visual verification of model stability.

cd backend
python -m venv venv
source venv/bin/activate  # or venv\Scripts\activate on Windows
pip install -r requirements.txt
python main.py

cd frontend
npm install
npm run dev

Primary application endpoint: http://localhost:3000

Proprietary technical demonstration. Optimized for mathematical precision and industrial-grade transparency.

D-Cipher AI: Decoding the Black Box.

This project is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). For commercial use, enterprise licensing, or closed-source integration (such as cloud deployment without open-sourcing your backend), please contact the author to arrange a commercial license.