Description:

This notebook creates an AI agent (DefenceAgent) to analyze text input or uploaded files for defence insights. The extension of this project would be to create an AI Defence Agent to analyse military documents.

Techniques Used:

Structured Output (JSON mode), Few-shot Prompting, Retrieval Augmented Generation (RAG), AI agent.

Input:

Text input or file (e.g., text file, CSV) containing defence-related data Note: Optimized for Kaggle with error handling and minimal dependencies.

Project roadblock:

Despite the code working, I kept getting errors when loading data from Kaggle or uploading data. So I had to use dummy data instead. In theory, this can be applied to other Kaggle Notebooks.

Example of a military defence text document:

Training model to assess threat levels

DefenceAgent class

End result is an AI Defence Agent that analyses reports to assess threat and confidence levels.

Techniques:

Retrieval Augmented Generation (RAG): Use TF-IDF cosine similarity to retrieve relevant defense profiles (avoids PyTorch/FAISS to avoid errors). It modifies interactions with a large language model so that the model responds to user queries with reference to a specified set of documents.

Few-shot Prompting: Few-shot prompting is a technique in artificial intelligence where a model is given a small number of examples to learn from before generating a response. This method helps improve the model's performance on specific tasks by guiding it with relevant demonstrations.

Structured Output (JSON mode): The final output is a structured JSON file with each military report assessed for threat and confidence levels in reported incidents on patrol.

Conclusion

This Kaggle Notebook demonstrated a few AI techniques that can be used to analyse military documents and be implemented by a data engineer in different circumstances.