// Hacker Noon · 11 March 2026

FogAI Part 3: The Knowledge Extraction Layer (Why Using an LLM for NER is Architectural Malpractice)

FogAi uses a Bi-Encoder Architecture to split the encoding process down the middle. It uses a single Python wrapper to execute MNN inference directly. FogAi leverages JNI and gRPC to execute inference directly without Python.

Hacker Noon

@hacker-noon · Nick Maletsky

hackernoon.com

Hacker Noon@hacker-noon