Neural Nexus 6

Nexus Analytics with Phara is a conversational web application designed to transform messy, real-world healthcare data into privacy-safe synthetic datasets through an agentic chatbot interface. Built to address the fragmentation, schema drift, and compliance risks inherent in payer and claims data, the system enables users to upload ZIP files containing healthcare records and receive a downloadable bundle of cleaned, redacted, and statistically faithful synthetic data. The backend pipeline performs schema inference, normalization, PII/PHI detection and masking, synthetic data generation with correlation preservation, and validation via delta reports comparing distributions and constraint violations. The frontend, served via Flask, connects directly to these backend endpoints, offering a zip-in/zip-out experience with a results dashboard and chatbot guidance. Technologies include Python 3.11, Flask, pandas, scikit-learn, presidio-analyzer, and multi-AI integration via OpenAI, Anthropic, Google Generative AI, and Ollama. User research involved interviews with healthcare analysts and ML engineers, whose feedback shaped the system’s emphasis on reproducibility, privacy rigor, and usability. The project meets all Optura Hackathon judging criteria, with seeded synthetic generation, modular architecture, and a merged requirements file enabling unified deployment. Key next steps include adding differential privacy toggles, expanding validation metrics, and refining frontend visualizations.