Minions

Minions is a communication protocol that enables small on-device models to collaborate with frontier models in the cloud. By only reading long contexts locally, we can reduce cloud costs with minimal or no quality degradation.

https://github.com/HazyResearch/minions

AIBrix

Cost-efficient and pluggable Infrastructure components for GenAI inference.

https://github.com/vllm-project/aibrix

Craw4LLM

CRAW4LLM is an efficient web crawling method that prioritizes webpages based on their potential influence on LLM pretraining, replacing traditional graph-connectivity-based priorities. By crawling only 21% of URLs, it achieves the same downstream performance as previous methods, significantly reducing data waste and website burden.

https://github.com/cxcscmu/Craw4LLM

How Python Evolves: From PEP to Feature

Explore how Python gets new functionality and how PEPs have a huge role in this, by looking at PEP484 and its interesting backstory.

https://www.youtube.com/watch?v=TzpOdpdX7pE

CCTV_YOLO

Fast Real-time Object Detection with High-Res Output

https://github.com/SanshruthR/CCTV_YOLO

logis

Turn your git commit history into a scientific log.

https://github.com/flywhl/logis

opendbc

A Python API for your car.

https://github.com/commaai/opendbc

AI Engineering Goes Visual: Web Scraping & Data Prep with PyFlyde

In part one of two part tutorial, we explore how to use Flyde, a visual programming tool, to build a web scraper that feeds into a Retrieval Augmentation Generation (RAG) system. We will cover the process of scraping web content and storing it locally, setting the stage for more advanced AI engineering tasks.

https://blog.kodigy.com/post/visual-ai-engineering-with-pyflyde-pt1-scraper/

NotaGen

NotaGen is a symbolic music generation model leveraging Large Language Models (LLMs) through pre-training on 1.6M musical pieces, fine-tuning on classical compositions, and reinforcement learning using a novel CLaMP-DPO method.

https://github.com/ElectricAlexis/NotaGen

The features of Python's help() function

Python has a built-in help function for getting help... but what can do with help?

https://www.pythonmorsels.com/help-features/

olmOCR

A toolkit for training language models to work with PDF documents in the wild.

https://github.com/allenai/olmocr

How I Automated My Podcast Transcript Production With Local AI

The author automated podcast transcription using roboscribe, a Python tool that combines WhisperX for diarized transcription and a local Large Language Model (LLM) for cleaning up the transcript, significantly improving readability. By leveraging local AI models, the author maintains control and optimizes the transcription process on their own hardware, achieving high-quality results in ...

https://den.dev/blog/how-i-automated-podcast-transcription-with-local-ai/

MLX-Audio

A text-to-speech (TTS) and Speech-to-Speech (STS) library built on Apple's MLX framework, providing efficient speech synthesis on Apple Silicon.

https://github.com/Blaizzy/mlx-audio

VoiceRestore

A cutting-edge speech restoration model designed to significantly enhance the quality of degraded voice recordings.

https://github.com/skirdey/voicerestore

academic-pretraining

$100K or 100 Days: Trade-offs when Pre-Training with Academic Resources.

https://github.com/apoorvkh/academic-pretraining