AI Agents · VI · Summary
Từ mô hình ngôn ngữ lớn đến tác tử AI tự hành
A practical distinction between traditional chatbots, LLM chatbots, RAG and AI agents, including the components that let an agent perceive, reason and act.
Read articleNUVERXAI TECH NOTES
Research, shared.
Engineering perspectives.
Technical writing on Edge AI, AI Agents, language models, speech and the path from research to product.
8 articles · Shown in their original language
An End-to-End ML Pipeline with Scikit-Learn, Pandas and MLflow
A guide to structuring a reproducible machine-learning pipeline, from data preparation to experiment tracking and model versioning.
01Fine-Tuning Pre-Trained Models for Computer Vision
A practical workflow for selecting a backbone, preparing data, freezing layers and evaluating a fine-tuned computer-vision model.
02Understanding Quantization for Large Language Models
An introduction to linear quantization, PTQ, QAT and common techniques for bringing language models to constrained hardware.
03Building a Large Language Model from Scratch
A guided overview from tokenization, embeddings and causal attention to the Transformer architecture and a small language-model training loop.
04A RAG Architecture for Large Document Collections
A proposed architecture for scaling RAG to hundreds of thousands of documents: structure-aware chunking, hybrid retrieval, quantized vector storage, semantic caching, cited generation and how to measure quality.
05Running Neural Networks on Raspberry Pi with TensorFlow Lite
An introduction to converting and running neural networks on Raspberry Pi with TensorFlow Lite, including what to measure on real hardware.
06Offline Speech Recognition and Synthesis on Raspberry Pi
A local speech-to-text and text-to-speech pipeline using Vosk and Piper, relevant to privacy-conscious on-device voice interaction.
07