The primary challenge we encountered was the lack of a suitable dataset specifically designed for detecting drug trafficking activities on social media platforms. This limitation made it difficult to train models directly for the task at hand, leading us to use publicly available Large Language Models (LLMs) as an alternative approach for detecting drug-related content. The variability in content types—ranging from text to images and videos—also posed a challenge, as it required integrating multiple AI models to analyze different formats effectively and identify trafficking patterns accurately.

Another significant hurdle was achieving real-time processing for analytics and alerting. Ensuring that the system could handle live data streams and provide timely insights required optimizing the backend with Node.js and leveraging MongoDB’s efficient querying capabilities for fast data retrieval and storage. These solutions helped us address the challenges, though the process highlighted the need for continuous refinement and adaptive strategies.