Sparrow Creator: Open-Source AI Doc Extraction ๐ | ML/Oracle Dev | @katana_ml | Try: sparrow.katanaml.io | github.com/katanaml
Joined March 2010
- Tweets 5,433
- Following 154
- Followers 6,688
- Likes 12,319
859 Photos and videos
Pinned Tweet
11 Nov 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
๐๐๐
11 Nov 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
1
6
22
Jun 13
The US just blocked non-Americans from accessing Anthropic's Fable 5 and Mythos 5. Overnight. No warning. ๐
This is exactly why I built Sparrow with local inference at its core โ your AI pipeline shouldn't depend on a political decision made thousands of miles away.
โก Running Mistral, Qwen or GLM, etc. locally with MLX-VLM or vLLM means zero exposure to these decisions.
๐ Document goes in โ structured JSON comes out. On your hardware. Offline. Always available.
๐ Own your models. Own your data. Run everything locally.
๐ sparrow.katanaml.io
๐ github.com/katanaml/sparrow
2
3
236
Andrej Baranovskij retweeted
Jun 13
Anthropic pulling the plug on Fable and Mythos 5 at the direction of a government order is a massive wake-up call. ๐ฅ
It highlights exactly why open, distributed AI matters.
When AI is centralized, access can disappear overnight.
Local models are different.
No one can revoke your access.
No one can pull the plug.
No one can take away the weights you already own.
Own your AI.
Jun 13
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-mytโฆ
24
38
312
11,952
Andrej Baranovskij retweeted
Jun 12
The video from @angeloskath on local agentic AI with MLX is excellent. I also hear it's one of the most viewed videos in WWDC history ๐
Goes through the basics of agentic AI and how to set it all up to run locally in a very approachable and simple way.
The demos are excellent and it's kind of wild that they just work now. None of this was possible or practical < 1 year ago before M5 and the recent quality bump in open weights models. And it's not done improving.
9
17
182
8,973
Jun 11
Integrated @MistralAI OCR as a cloud backend into Sparrow โ open-source document extraction platform (5.2k โญ)
Two-step pipeline:
Mistral OCR โ structured HTML
Mistral Small โ JSON extraction
Works alongside local backends (MLX, vLLM) โ same API, just a flag switch
Full local or full cloud. Enterprise covers both.
โญ github.com/katanaml/sparrow
@MistralDevs @sophiamyang
1
4
163
Jun 10
Gemma 4 12B vs Ministral 14B โ structured table extraction, JSON schema, array output.
โ Gemma 4 12B (8-bit and bf16): fails to return a proper JSON array
โ
Ministral 14B 8-bit: extracts all rows correctly
Tested with Sparrow ๐
๐ฌ youtube.com/watch?v=4yFW_mmzโฆ
๐ sparrow.katanaml.io
1
1
5
186
Andrej Baranovskij retweeted
Jun 9
One of my personal favorite features announced at WWDC will I suspect be a sleeper hit: container machines, allowing your Mac to run a lightweight, persistent Linux environment with your home directory and repos automatically mounted: github.com/apple/container/bโฆ
227
815
9,698
728,596
Watch this year's WWDC video for MLX!
youtube.com/watch?v=wykPErJ8โฆ
4
17
83
15,979
Andrej Baranovskij retweeted
Jun 9
Three MLX videos dropped at WWDC:
Running agents locally by @angeloskath youtube.com/watch?v=wykPErJ8โฆ
Distributed inference and training by Tatiana Likhomanenko youtube.com/watch?v=CzgK02zsโฆ
MLX Swift by David Koski youtube.com/watch?v=KCL8f9ztโฆ
10
43
363
40,268
Andrej Baranovskij retweeted
Jun 7
Cybertruck sunrise on 280 right now headed to San Francisco.
14
5
150
17,304
Jun 4
New Sparrow UI vibe coded with Claude Code is live in prod: sparrow.katanaml.io
Shadcn and Next.js stack
1
1
4
174
Andrej Baranovskij retweeted
Jun 3
Meet Gemma 4 12B!
A unified, encoder-free multimodal model designed to bring high-performance intelligence directly to your laptop, and released under an Apache 2.0 license.
Bridging the gap between edge efficiency and advanced reasoning. Here is whatโs new with Gemma 4 12B: ๐
404
1,789
12,365
3,176,260
Jun 2
New Sparrow UI for mobile will be much better, thanks to Claude Design
1
4
222
Jun 2
Prince is doing incredible job and pushing new features for mlx-vlm. This is super useful, thanks ๐
Jun 1
Today we're shipping our biggest MLX-VLM release yet: v0.6.0
...and we are raising ๐ธ
This one's about turning your Apple devices into real local agent machines. From your desk to your pocket.
What's new:
โก Speculative decoding everywhere โ Gemma 4 EAGLE3 DFlash, Qwen MTP, DeepSeek V4 MTP. Faster tokens, less waiting.
๐ค Agent-ready server โ native Anthropic /v1/messages API, stateful /v1/responses, tool calls, Codex context budgets. Plug Claude Code & Codex straight into local models.
๐๏ธ New models galore โ DeepSeek V4, ZAYA1-VL, MiniCPM-V 4.6, LFM2 MoE, Step-3.7 Flash, Laguna more.
๐จ Image gen & editing โ FLUX.2 (base klein), PrismML Bonsai.
๐ Audio in โ Qwen3 Omni, Gemma 4 audio, base64 chat audio.
๐งฎ TurboQuant KV cache โ RHT-correct fast paths for leaner memory.
๐ฆ Modular server, better metrics, cleaner streaming.
Run real agents on the hardware already in your hands.
Github: github.com/Blaizzy/mlx-vlm
3
289
Jun 1
Building Agentic AI Pipelines for Document Analysis
Two steps. Fully local.
1๏ธโฃ Sparrow Parse extracts structured data from bonds table โ Ministral 3B 14B
2๏ธโฃ Sparrow Instructor analyzes portfolio risk โ Gemma 4 31B
Orchestrated with Prefect. No data leaves your machine.
YouTube: youtube.com/watch?v=Sw_zzzu7โฆ
GitHub: github.com/katanaml/sparrow
Sparrow: sparrow.katanaml.io
2
189
May 31
New Sparrow UI with shadcn/ui and next.js is coming up. Functionality is working, entering testing/fixing phase.
1
213
May 28
Building new Sparrow UI with Claude Code is going well. Implemented file upload component, added backend code with Next.js
Migrating to shadcn from Sparrow Gradio UI:
sparrow.katanaml.io
1
1
2
377