46 lines (34 loc) · 1.47 KB

Use Cases

This folder contains practical ML examples that sit closer to applied workflows than the learn/ directory.

Best Starting Points

learning_with_less/
- Runnable locally.
- Covers supervised, self-supervised, and semi-supervised learning with an offline fallback dataset.
SparkImageProcessing/
- Runnable locally if you have OpenCV and optionally Spark.
- Good for understanding the difference between a simple local pipeline and a Spark-style batch pipeline.
RealTimeDataProcessing/
- Runnable if you have Kafka and Spark installed.
- Shows the shape of a streaming fraud-detection workflow.

Requires External Services Or Keys

Agents/
- LLM agent demos.
- Typically require MGA_API_KEY and the relevant Python packages.
LLMs/
- DSPy and research-generation examples.
- Typically require OPENROUTER_API_KEY, and some scripts also require TAVILY_API_KEY.

Cloud Or Notebook Heavy Examples

Sagmeaker_Fraud_Detection_End_To_End/
- SageMaker notebook workflow.
demand_prediction/
- Notebook examples.

Media-Specific Example

sportanalytics/
- YOLO-based sports video tracking demo.
- Requires a local video file and the provided yolo11n.pt weights.

Working Rule

If you want examples that are easiest to study from first principles, start with the folders above that now have:

clear READMEs
relative paths instead of machine-specific paths
import-safe Python scripts
local fallbacks where possible