Predicting local weather patterns (1–6 hours ahead) is computationally expensive and difficult with traditional numerical models, especially at local spatial resolution.
Developed a ConvLSTM deep learning model to predict the next 6–60 frames of radar echo images based on historical sequences from the CIKM radar dataset.
Related case studies
96%
Hybrid CNN-RNN model for Urdu sentiment classification, published in a peer-reviewed journal with 96% F1 score.
97–98%
Agentic RAG legal assistant for Saudi regulatory law. 97–98% recall@10 across a bilingual English-Arabic corpus validated by legal domain experts.
90%+
Agentic RAG platform indexing 300+ GB of petroleum engineering documentation. 90%+ retrieval hit rate on domain-heavy technical queries.
