DataChain Blog
Find here DataChain and DVC news, findings, interesting reads, community takeaways, deep dive into machine learning workflows from data versioning and processing to model productionization.
Parquet Is Great for Tables, Terrible for Video - Here's Why
Parquet is great for tables, terrible for images and video. Here's why shoving heavy data into columnar formats is the wrong approach - and what we should build instead. Hint: it's not about the formats, it's about the metadata.
From Big Data to Heavy Data: Rethinking the AI Stack
LLMs can finally interpret unstructured video, audio, and documents — but they can't do it alone. This post introduces the concept of heavy data and explores how modern teams build multimodal pipelines to turn it into AI-ready data.
As GenAI Fever Fades - Time to Prioritize Robust Engineering Over Overblown Promises
Improved Engineering and Data Management will be what carries GenAI into maturity
Scalable PDF Document Processing with DataChain and Unstructured.io
Extract and parse text from documents and create vector embeddings in a scalable and distributed way (and less than 70 lines of code).
Post-modern AI Data Stack
How and Why Generative AI will change the modern data stack.
You Do the Math: Fine Tuning Multimodal Models (CLIP) to Match Cartoon Images to Joke Captions
Learn how to fine tune multimodal models like CLIP to match images to text captions.
Enforcing JSON Outputs in Commercial LLMs
The results of our tests on the structured outputs of Google Gemini Pro, Anthropic Claude, and OpenAI GPT. DataChain used for evaluation.