<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Ridhwan Amin — AI/ML Researcher</title><description>AI/ML researcher at UTP. I build seismic inpainting models, anomaly detection systems, and hybrid RAG pipelines.</description><link>https://ridhwan.dev/</link><item><title>Why Hybrid Retrieval Beats Dense-Only RAG in Domain-Specific Tasks</title><link>https://ridhwan.dev/posts/why-hybrid-retrieval-beats-dense-only/</link><guid isPermaLink="true">https://ridhwan.dev/posts/why-hybrid-retrieval-beats-dense-only/</guid><description>Dense embeddings miss exact terminology. Here&apos;s why combining BM25 with vector search improved our petroleum Q&amp;A system by 12 percentage points — and how to implement it.</description><pubDate>Thu, 18 Jan 2024 00:00:00 GMT</pubDate></item></channel></rss>