Nads | Devpost

Inspiration

Anomaly detection is critical across domains like infrastructure monitoring, cybersecurity, finance, and IoT but many existing tools are either too complex to set up or too opaque to interpret. We wanted to build a system that makes anomaly detection immediate, visual, and accessible, without sacrificing technical depth. NADS was inspired by the idea that AI systems should not only detect problems, but also help humans understand them quickly.

What it does

NADS is a real-time AI-powered anomaly detection platform that allows users to upload datasets, automatically identify anomaly-relevant patterns, and monitor abnormal behavior through an interactive dashboard. It supports multiple data formats and anomaly detection algorithms, providing visual insights through charts, heatmaps, and live monitoring simulations.

How we built it

We built NADS using Python and Streamlit for rapid development and cloud deployment. Data processing is handled with Pandas and NumPy, while anomaly detection is performed using machine learning models such as KMeans, DBSCAN, Isolation Forest, and an LSTM-based simulation. Custom CSS and animations were used to create a neon cyberpunk interface optimized for clarity and responsiveness.

Challenges we ran into

One major challenge was balancing performance with an animation-heavy interface in Streamlit. We also had to design a flexible data ingestion pipeline that works across diverse datasets without manual configuration. Simulating real-time monitoring in a static data environment required careful state management and optimization.

Accomplishments that we're proud of

Built a fully functional anomaly detection dashboard within hackathon constraints Integrated multiple anomaly detection models into a single workflow Designed a visually distinctive yet usable cyberpunk-inspired interface Deployed the application in a cloud-ready, modular architecture

What we learned

We learned how to combine machine learning models with interactive visualization to improve interpretability. The project deepened our understanding of anomaly detection techniques, Streamlit performance optimization, and the importance of user-centered AI design.

What's next for Nads

Next, we plan to add support for live data streams, improve explainability of detected anomalies, and integrate real alerting mechanisms such as email or webhook notifications. We also aim to adapt NADS for real-world use cases like cybersecurity monitoring and industrial systems.