Andrej Baranovskij

@andrejusb

Sparrow Creator: Open-Source AI Doc Extraction 🚀 | ML/Oracle Dev | @katana_ml | Try: sparrow.katanaml.io | github.com/katanaml

Katana ML 👉

katanaml.io

Joined March 2010

Replies Articles Media

Pinned
Andrej Baranovskij
@andrejusb
Nov 11, 2021
We launched Katana ML katanaml.io in 2018 and now it is time to update the website, to explain where we are now and what we do with #MachineLearning, #MLOps, and #opensource 🚀🚀🚀
Katana
@katana_ml
Nov 11, 2021
We have a new website - katanaml.io It explains what we do with ML in a simple and straightforward way. It is featuring our open source product Skipper, we are using it to run #MLOps. #MachineLearning #MLOps
Andrej Baranovskij
@andrejusb
Aug 10
🤔 Bigger model ≠ better extraction. Same insurance pivot table, same "*" query, no schema — but this time Sparrow's Advanced mode (Gemma 4 31B) vs Standard mode (Ministral 14B). 📉 Gemma 31B, 2x the size, was LESS accurate. 🧩 Hallucinated on columns — nulled real values,
Andrej Baranovskij
@andrejusb
Aug 5
🐦 Can an LLM read a pivot table? Pivot tables are a genuinely hard case for extraction — row/column headers, merged cells, aggregated values instead of a flat list. Tested Ministral 14B in Sparrow Standard mode on an insurance pivot table. 1️⃣ Query: generic "*" — no
Andrej Baranovskij
@andrejusb
Jul 26
Article
Sparrow Standard Mode: Document Extraction with Ministral 14B
Sparrow's Standard tier now runs on Ministral 3 14B via vLLM. Fast, local, no cloud dependencies. Tested through Sparrow's UI at sparrow.katanaml.io/process. Query: Result: Three records, correctly...
Andrej Baranovskij
@andrejusb
Jul 22
Sparrow's Standard mode in action — Ministral 14B extracting structured document data into clean JSON. Fast, accurate, fully local. No cloud dependencies. sparrow.katanaml.io youtube.com/watch?v=GCJ_l1…