NVIDIA DGX Spark Hackathon

Build a system that derives actionable insights from Austin's real-time traffic incident reports to help predict high-risk corridors and position safety assets proactively.

Goal: Predictive alerting, contextual intelligence correlating weather with crashes, or digital twin simulations.

Dataset: Austin Real-Time Traffic Incident Reports

Process egocentric (POV) video in real-time to generate safety interventions or operational knowledge in industrial environments.

Goal: Real-time safety violation detection, automated SOP documentation, or ergonomic analysis.

Dataset: Egocentric-10K (POV Video of industrial tasks)

Unlock unstructured permit data to reveal the city's physical future or economic trends from 2M+ rows of Austin construction permits.

Goal: Hidden trend discovery, development advisor RAG agent, or economic forecasting.

Dataset: Issued Construction Permits (Austin)

Build a forecasting and optimization system for U.S. multimodal freight pipeline, predicting congestion and recommending ideal rail dispatch windows.

Goal: 24-72 hour port surge predictions, rail congestion warnings, and multimodal flow visualization.

Dataset: PortWatch, North American Rail Network, County-to-County Truck Travel Times, Logistics Fleet Data

Measure ML's real influence on scientific progress across disciplines using 81,000 scientific papers spanning 2016-2024.

Goal: Interactive research intelligence platform quantifying ML impact, adoption dynamics, and quality trade-offs.

Dataset: 81,000 scientific papers with citation networks, ML adoption markers, and reproducibility indicators

Train a model that transforms unedited real estate photos into production-ready, professionally edited images.

Goal: 70% image quality (matching professional edits) + 30% inference cost efficiency.

Dataset: 577 paired images (input → professionally edited output)