Rajistics blog
I have been skeptical of Graph RAG. I made a short video about that skepticism , and my issue was not that graphs are useless. The issue was the plumbing. For a lot of business problems, you have to decide what the nodes are, what the edges are, how to keep the graph updated, and how the agent is supposed to query it. By the time that is working, it is easy to wonder whether you built…
I keep hearing the model labs tell a story that as models get better, you need to give the model less direction, and the coding agent harnesses around them should get simpler. I like that story, it sounds so intuitive. But like a good engineer, I also do not quite trust it. So I pulled together a small visual experiment around Claude’s published system prompts. The original version looked at Opus…
I sit in the middle of corn fields and watch the live stream. Thanks @aiDotEngineer and @swyx. Here are some talks from AI Engineer World’s Fair 2026 that I found particularly interesting, in order of presentation, not quality. AI Engineer World’s Fair 2026 keynotes Day 1 Microsoft Foundry: Knowledge Management for AI Agents Watch the talk URL: https://www.youtube.com/watch?v=htM02KMNZnk&t=1800s…
Coding agents are forcing a shift in how we think about AI systems. A few years ago, many of us were using language models for short, stateless tasks. In this talk, Rajiv walks through why long-running coding tasks create a different engineering problem and why the system around the model now matters as much as the model itself. This version follows the exported deck page by page, using the PDF…
I’ve always been interested in how teams innovate when building AI. At AI Engineer Europe 2026 , the talks that stayed with me weren’t the ones about bigger models or faster demos. They were the ones where teams hit a wall, realized their first approach wasn’t working, and changed the architecture. These reframing examples stood out to me. They’re stories about changing the unit of work, changing…
Annoyed by the “X is dead” or “have you seen this game-changer technology?” posts? With algorithmic feeds, this kind of content bubbles to the top on X and LinkedIn. I found this super frustrating. I am looking for the latest real news or deeper, substantive information. The platforms are optimizing for clickbait type views and for me the signal-to-noise ratio on these platforms was unbearable. A…
target_leakage.jpg Machine learning is a glass cannon. When used correctly, it can be a truly transformative technology, but just a small oversight can cause it to become misleading and even actively harmful. Even if all the code runs and the model seems to be spitting out reasonable answers, it’s possible for a model to encode fundamental data science mistakes that invalidate its results. These…
Generative AI projects rarely fail because teams cannot get a model to produce a good demo. They fail because teams cannot reliably tell whether the system is improving, regressing, or creating new business risk. A single strong response can hide dozens of subtle failures in production. This guide is my practical workflow for evaluating Generative AI applications once you move past the “vibe…
Video Watch the full video Annotated Presentation Below is an annotated version of the presentation, with timestamped links to the relevant parts of the video for each slide. Here is the slide-by-slide annotated presentation based on the video “From Vectors to Agents: Managing RAG in an Agentic World” by Rajiv Shah. This was presented in different forms at several conferences in the Fall of 2025…
Want to know the best places for the latest data science, AI, ML news? Here’s my filtered news feed (Updated February 2026). AI/ML is too huge to try and capture all the news sources. I try to keep a diverse range to keep my pulse on what’s new and useful. Most great data scientists I know probably spend an hour a week reading the news, so please don’t focus on consuming content (build…
When working with machine learning models, preparing data properly is essential. One common preprocessing technique is one-hot encoding, which transforms categorical data into a format algorithms can understand. However, this transformation often creates sparse matrices - dataframes where most values are zero. Basic One-Hot Encoding The first animation illustrates the fundamental concept of…
How to Select the Best Features for Machine Learning! Let’s deep dive into several feature selection techniques and help you figure out when to use each one. The notebook includes two data sources: the MNIST dataset and the Madelon dataset. The MNIST dataset is a collection of 28x28 pixel images of handwritten digits. The Madelon dataset is a synthetic dataset that you can control. The notebook…
Video Watch the full video Annotated Presentation Below is an annotated version of the presentation, with timestamped links to the relevant parts of the video for each slide. Here is the annotated presentation for “Rules: A Simple & Effective Machine Learning Approach” by Rajiv Shah. 1. Title Slide Slide 1 ( Timestamp: 00:00:00 ) The presentation begins by introducing the core topic: Interpretable…
Video Watch the full video Annotated Presentation Below is an annotated version of the presentation, with timestamped links to the relevant parts of the video for each slide. Here is the annotated presentation based on the provided video transcript and slide summaries. 1. The Spark of the AI Revolution Slide 1 ( Timestamp: 00:00 ) The presentation begins with the title slide, “The Spark of the AI…
Video Watch the full video Annotated Presentation Below is an annotated version of the presentation, with timestamped links to the relevant parts of the video for each slide. Here is the annotated presentation based on the video transcript and slide summaries. 1. Title Slide: A Practical Perspective on Generative AI Slide 1 ( Timestamp: 00:01 ) This presentation begins with an introduction by…
This notebook introduces several key features of Snowflake ML in the process of training a machine learning model for forecasting Chicago bus ridership. Establish secure connection to Snowflake Load features and target from Snowflake table into Snowpark DataFrame Prepare features for model training Train ML model using Snowpark ML distributed processing Save the model to the Snowflake Model…
Video Watch the full video Annotated Presentation Below is an annotated version of the presentation, with timestamped links to the relevant parts of the video for each slide. Here is the annotated presentation for “Evaluating LLMs” by Rajiv Shah. 1. Title Slide: Evaluating LLMs Slide 1 ( Timestamp: 00:00 ) The presentation begins with the title slide, introducing the speaker, Rajiv Shah, and the…
Reasoning Introduction I was wowed by ChatGPT. While I understood tasks like text generation and summarization, something was different with ChatGPT. When I looked at the literature, I saw this work exploring reasoning. Models reasoning, c’mon. As a very skeptical data scientist, that seemed far-fetched to me. But I had to explore. I came upon the Big Bench Benchmark , composed of more than 200…
SetFit Introduction We change our conversational style from informal to formal speech. We often do this without thinking when talking to our friends compared to addressing a judge. Computers now have this capability! I use textual style transfer in this post to convert informal text to formal text. To make this easy to use, we do it in a spreadsheet. Step 1 The first step is identifying an…
SetFit Introduction Data scientists often do not have large amounts of labeled data. This issue is even graver when dealing with problems with tens or hundreds of classes. The reality is very few text classification problems get to the point where adding more labeled data isn’t improving performance. SetFit offers a few-shot learning approach for text classification. The paper’s results show…