Aniss Djellal

A solution for the 3blue1brown's june 2026 puzzle

data-free pipeline that compresses LLMs to 3.125 bits per weight — no calibration data, no fine-tuning, no dequantization at inference

Why deterministic query translation should often come before embeddings in enterprise RAG systems, and how to combine both in a hybrid approach.