Articles - Page 2

Running Gemma 4 31B Faster at Home with Speculative Decoding

15-May-2026

Running Gemma 4 31B Faster at Home with Speculative Decoding

A practical guide to running Gemma 4 31B on a dual-GPU Ubuntu box using llama.cpp TurboQuant, Gemma 4’s MTP assistant head, and speculative decoding to nearly double token throughput.

How We Made RAG Indexing Faster With an Adaptive Embedding Endpoint Pool

11-Apr-2026

How We Made RAG Indexing Faster With an Adaptive Embedding Endpoint Pool

A simple explanation of how to speed up embeddings by routing work across fast and slow local AI endpoints without letting one slow batch block the whole indexing pipeline.

Why Vector Search Is Harder Than It Looks (And Why It Matters)

10-Apr-2026

Why Vector Search Is Harder Than It Looks (And Why It Matters)

A simple, practical introduction to embeddings, vector indexes, and real-world semantic search

Gemma 4 Explained: Google's Open-Source AI That Runs on Your Phone

05-Apr-2026

Gemma 4 Explained: Google's Open-Source AI That Runs on Your Phone

A comprehensive, accessible guide to Google Gemma 4's architecture, multimodal capabilities, Mixture of Experts, Per-Layer Embeddings, and real-world deployment on phones, laptops, and servers.