Harsh Parikh

The project aimed to develop a disaster management solution that could ingest and analyze real-time data. Due to varying data formats, we faced a major setback in integrating data from diverse sources such as satellites, social streams, webcams, and OSINT. To overcome this hurdle, I developed a dynamic data pipeline that leveraged metadata like geotag and timestamps to seamlessly integrate all data sources, enabling the inference engine of LLM to generate text and image embeddings for real-time analysis.