Machine Learning
Vision LLMs are PDF Parsers Too: Reading Charts and Diagrams for RAG
3 weeks ago
Vision LLMs are PDF Parsers Too: Reading Charts and Diagrams for RAG
companion in Enterprise Document Intelligence, the series that builds an enterprise RAG system from four bricks. Article 5 (document parsing)…
GPU Time-Slicing for Concurrent LLM Agents on Kubernetes
3 weeks ago
GPU Time-Slicing for Concurrent LLM Agents on Kubernetes
. Production agents fight over the same GPU — and on one shared card, a latency-sensitive agent’s p99 latency quietly…
Larger Context Windows Don’t Fix RAG — So I Built a System That Does
3 weeks ago
Larger Context Windows Don’t Fix RAG — So I Built a System That Does
TL;DR I built a dataset Q&A system and trusted a RAG answer that was less than half-correct. I measured this…
Parse PDFs for RAG Locally with Docling: Rich Tables, No Cloud Upload
3 weeks ago
Parse PDFs for RAG Locally with Docling: Rich Tables, No Cloud Upload
companion in Enterprise Document Intelligence, the series that builds an enterprise RAG system from four bricks. Article 5 (document parsing)…
Solving the 3Blue1Brown String Probability Problem (Without AI)
3 weeks ago
Solving the 3Blue1Brown String Probability Problem (Without AI)
a stupid chance problem solving problem in my free time when I could be scrolling through doom? Because I'm trying…
When PyMuPDF Can’t See the Table: Parse PDFs for RAG with Azure Layout
3 weeks ago
When PyMuPDF Can’t See the Table: Parse PDFs for RAG with Azure Layout
companion in Enterprise Document Intelligence, the series that builds an enterprise RAG system from four bricks. Article 5 (document parsing)…
Why Decade-Old Residual Connections Still Power All of AI (And Why That’s a Problem)
3 weeks ago
Why Decade-Old Residual Connections Still Power All of AI (And Why That’s a Problem)
1. Over the past decade, deep learning as a field has grown quite significantly, whether it be the compute capacity…
A Harness for Every Task: Putting a Team of Claudes on One Job
3 weeks ago
A Harness for Every Task: Putting a Team of Claudes on One Job
1. For most of 2024 and 2025, the default answer was simple: give the task to one agent, use the…
I thought Data Engineering was just Writing Documentation. I was wrong.
3 weeks ago
I thought Data Engineering was just Writing Documentation. I was wrong.
The ETL pipeline, I thought I had a good grasp of what data engineering really is. You extract data from…
Is Language Recognized? Experimenting with Chinese characters
3 weeks ago
Is Language Recognized? Experimenting with Chinese characters
widely discussed on Douban – a Chinese social network – about a broken printer. The owner noticed that when the…