A detailed recap of the ODSC AI 2026 Day 3 sessions, covering keynotes, lectures, and workshops on topics such as trustworthy LLMs, clustering, reinforcement learning for…
A deep dive into Dr. Ali Arsanjani’s talk on architectural patterns for building and governing production-grade multi-agent systems.
A deep dive into building Claude-based agents for financial-services industry workflows, exploring best practices, examples, and hands-on practice.
A deep dive into Achyut Sarma Boggaram’s talk on deploying multimodal AI at the edge, focusing on engineering patterns for real-world performance.
A deep dive into Kyle Stratis’s talk on using the Model Context Protocol to support agentic behaviors in applications.
A deep dive into Ryan Lopopolo’s talk on Harness Engineering, exploring practical patterns for agent-first software development.
A deep dive into Kyle Stratis’s talk on AI Agents with MCP, exploring the practical implementation and best practices for building AI agents using the Model Context Protocol.
A deep dive into Janet Six’s talk on meaningful data visualization in the age of AI, exploring how to make visualizations that truly support decision-making and…
A deep dive into Ivan Lourenço Gomes’s talk on practical strategies for organization-wide AI adoption, exploring how to implement AI effectively across teams and workflows.
A deep dive into Deep Patel’s talk on real-time event-time consistent analytics pipelines, exploring the use of Kafka, Flink, and Apache Pinot for low-latency…
A deep dive into Pedro Domingos’s talk on Tensor Logic, exploring the unification of symbolic reasoning and neural computation in a single language for AI.
A deep dive into Andy Petrella’s talk on the operational transformation of data architecture, exploring the unification of symbolic reasoning and neural computation in a…
A deep dive into Ryan Day’s talk on Model Context Protocol (MCP), exploring how to provide data and functionality to AI agents using Python.
A deep dive into Harry Mellor’s talk on vLLM with the Transformers Modelling Backend, exploring the integration of vLLM with HuggingFace Transformers for efficient model…
A workshop on building agentic AI applications from scratch, covering tool use, task decomposition, self-correction, conditional routing, and human-in-the-loop review.
A workshop on building responsible AI agents using open source tools, with a focus on healthcare applications and privacy-preserving design.
In this funky workshop, Rajiv Shah dives into the rabbit hole of harness engineering for AI agents. He covers the importance of coding harnesses, the key levers of harness…
An introduction to machine learning concepts and practical application using scikit-learn, covering topics such as regression, classification, decision trees, random…
A workshop introducing the Agent-to-Agent (A2A) protocol, which enables interoperability and communication between independent AI agents across different frameworks and…
An accessible introduction to the mathematics behind transformers and large language models, covering key concepts such as attention mechanisms, self-attention, multi-headed…
These are workshops and sessions that took place before the start of the main track.
An exploration of the challenges associated with mixture models in Bayesian statistics, including identifiability, label switching, and computational complexity, along with…
In this talk, Valentin Vercovici and Callan Fox discuss the challenges of maintaining high KV cache hit rates in large-scale AI inference workloads and introduce the concept…
In this talk, Chaim Rand revisits the NVIDIA Nsight profiling tools to augment the PyTorch Profiler, addressing critical issues in data transfer and batching for AI/ML…
Exploring how Reinforcement Learning (RL) can enhance the reasoning and planning capabilities of Large Language Models (LLMs) by creating a hierarchical agent structure and…
A deep dive into the three modes of convergence for random variables: convergence in probability, almost sure convergence, and convergence in distribution. We explore their…
This is my guide for sorting out the fluff from the substance around AI agents. I want to track what people are most excited about so I can join in the fun.
Convolutions are a fundamental mathematical operation that arises in many areas of mathematics, physics, and engineering. In this post, we will explore the concept of…
I’ve been sick for about a week with the sniffles and a mild fever. It’s been a bit of a rough patch, but I’m starting to feel better now. I’ve been resting a lot and…
A detailed recap of the PyData Global 2025 talk on Anthropic’s Claude, covering its architecture, safety features, and applications in AI.
A comprehensive recap of the PyData Global 2025 talk on integrating causal thinking into machine learning, focusing on uplift modeling to enhance decision-making processes…
A detailed recap of Scott Routledge talk at the PyData Global 2025 conference. In which he discusses Bodo DataFrames, a high-performance computing-based drop-in replacement…
A detailed recap of the PyData Global 2025 talk on combining Zarr, HDF5, and TIFF into a single data format, covering key components, performance considerations, and…
A detailed recap of the PyData Global 2025 talk on GPU-accelerated Zarr, covering key components, performance considerations, and practical implementation tips.
A detailed recap of the PyData Global 2025 talk on GPU-accelerated Python using RAPIDS, covering key components, common pitfalls, and practical implementation tips.
A recap of Itai Gilo’s PyData Global 2025 talk on defensible, reproducible, and auditable machine learning pipelines, with a focus on data versioning and compliance controls.
A detailed recap of the PyData Global 2025 talk by Claudio Salvatore Arcidiacono on effectively using text embeddings in tree-based models, covering key concepts…
A detailed recap of the PyData Global 2025 talk on extending sparse computing in the Python ecosystem using PyData/Sparse and the Finch compiler, covering current landscape…
A detailed recap of the PyData Global 2025 talk on building a transformer model from scratch using PyTorch, covering key components, common pitfalls, and practical…
A detailed recap of the PyData Global 2025 talk on lessons from the Chernobyl disaster applied to machine learning systems in production, highlighting common failure modes…
A live demonstration of how PyCaret simplifies machine learning workflows by enabling users to train and compare multiple models with minimal code.
An exploration of how Bayesian statistical modeling can be applied to the cold start problem in recommendation systems, enabling efficient decision-making and risk…
A recap of Jayita Bhattacharyya’s talk at PyData Global 2025 on Small Language Models (SLMs), exploring their capabilities, training techniques, and implications for AI…
A practical introduction to Bayesian decision theory, illustrating how to make optimal decisions under uncertainty with hands-on Python examples.
A practical tutorial on using Blosc2 to optimize data workflows in Python, focusing on out-of-core computations and efficient handling of large datasets.
A practical guide to optimal variable binning techniques for enhancing logistic regression models in regulated industries.
A comprehensive framework for monitoring, analyzing, and optimizing AI/ML workloads to enhance resource utilization and cost attribution using Metaflow.
A recap of the PyData Global 2025 talk on real-time financial fraud detection using modern Python tools and techniques.
A practical introduction to survival analysis, its core concepts, and applications using Python and R.
A hands-on tutorial on time series analysis of coupled neuron models using Python.
A recap of Adam Cowley’s PyData Global 2025 talk on using Model Context Protocol (MCP) to transform Anthropic’s Claude into a football opposition analyst.
A detailed recap of the PyData Global 2025 workshop on building LLM-powered applications, focusing on software engineering principles, prompt engineering, monitoring, and…
A hands-on walkthrough of building a fast, offline-capable reverse geocoder in Python using open-source tools like cKDTree, shapely, and geopandas.
Exploring the transition from traditional feature engineering techniques to context engineering for large language model agents.
An in-depth exploration of using various generative models to create synthetic non-life insurance premium data, including validation techniques and model comparisons.
A hands-on walkthrough of building the original Transformer architecture from scratch in PyTorch, demystifying attention mechanisms, positional encoding, and the…
A recap of the PyData Global 2025 talk by Daniel Chen on integrating LLMs into data science workflows, using the Chatlas package and Shiny dashboards to create AI-powered…
A detailed exploration of optimizing large-scale pandas applications using high-performance alternatives like Polars, FireDucks, and cuDF, highlighting best practices and…
A comprehensive overview of how Python can be effectively integrated with Excel to automate and enhance data workflows, focusing on practical tools and techniques for…
A humorous look at the worst practices in Python programming, presented by data and analytics comedian Evan Wimpey.
A practical guide to implementing BM25 for fuzzy product matching at scale using Python and DuckDB, complete with performance benchmarks and optimization tips.
A comprehensive overview of transitioning Jupyter notebooks from exploratory tools to production-grade pipelines, covering best practices, tools, and real-world challenges.
A comprehensive overview of how traditional AI and large language models can be leveraged to automate the creation of complex and critical documents in healthcare, such as…
In this talk, Aarti Jha presents a practical guide to understanding, detecting, and mitigating hallucinations in AI systems for building trustworthy applications.
Exploring the challenges and strategies for defining and measuring metrics in modern AI systems, especially when traditional metrics are unavailable or inadequate.
An in-depth look at the architectural differences that make Julia’s GPU-accelerated ODE solvers significantly faster than their Python counterparts, based on the 2024…
A detailed recap of the PyData Global 2025 talk on projspec, a tool for introspecting and managing Python data projects.
An in-depth look at how Insee modernized text classification by transitioning from fastText to a PyTorch-based model using the open-source torchTextClassifiers package…
A few thoughts on how to get the most out of GPT-5 for coding and tough problems
A model for marketing mix
A critical analysis of the allegations of war crimes and genocide in the context of the Israeli-Palestinian conflict, examining the evidence, legal standards, and the…
A collection of tools and techniques I use in the AI age.
This is my vibe coding guide on how I built a POC of concept for running Mesa app (their tutorial) in ShinyLive app so it can run in the browser without a front end.
A brief review of nonparametric statistical methods.
Exploring GenAI-powered tools
Some thoughts and code for a baseline morphology and syntax for emergent languages
A list of clarified research questions regarding emergent languages, planning, and grounding in multi-agent systems.
Thomas Schelling’s 1973 article explores binary choices where one person’s decision affects others, called by economists as externalities
How do could RL agents infer the goals of other agents?
Model for a cognitive mechanism similar to computer engines that simulate rich physics in video games and graphics, but that uses approximate, probabilistic simulations to…
Learning priors for Shapes
teaser for reading this paper
Zollman’s paper on how adding spatial structure affects the evolutionary outcomes of games with emergent communication and social cooperation
An exploration of partial pooling equilibria and mixed strategies in the Lewis signaling game, shedding light on their implications for emergent languages and communication…
A working paper on emergent languages
Some questions about planning in the Lewis Signaling Game and the stability of equilibria.
A review and study guide of the paper ‘On Learning To Become a Successful Loser: A Comparison of Alternative Abstractions of Learning Processes in the Loss Domain’ by…
Review of ‘Emergence of Linguistic Communication from Referential Games with Symbolic and Pixel Input’ by Angeliki Lazaridou et al.
A review of the paper ‘Linguistic generalization and compositionality in modern artificial neural networks’ by Marco Baroni.
Review of Mandelbrot’s papers on scaling in financial prices and his popular science book The (Mis)behaviour of Markets.
This article explores Vitter’s Algorithm and its application in online learning and communication within multi-agent systems.
Can we fine tune a state of the art LLM to improve the readability of text?
A review of Marco Baroni’s talk on the emergence of languages and the role of compositionality in language evolution.
In this post, I explore ideas for extending the Sugarscape environment to create a richer curriculum learning environment for multi-agent reinforcement learning, with a…
In this post, I review the paper ‘Multi-agent Reinforcement Learning in Sequential Social Dilemmas’ by Leibo et al. (2017), which introduces a new class of social dilemmas…
A deep dive into the complex signaling systems
In this post I share some of the lessons I learned from implementing Sugarscape in Mesa. I cover the design decisions I made and how they helped me to overcome some of the…
A quick note about Bayesian Agents and how to implement them using the MESA framework in Python.
How can we ensure signals are honest in a world where deception is rewarded? This paper delves into the theory of honest signalling in animal behavior, specifically…
Patrick Henry Winston provides a definition of AI. How he keeps expanding his definition as he goes along is thing of beauty. It expands organicaly to incorporate new ideas…
Some thoughts on the Lewis signaling game
Quarto at last supports Mindmap charts using Mermaid charts.
some migration notes from Blooger to Jekyl to Quarto blog.
SuperLearner is an ensambeleing library.
Autogluon is a auto-ml framework, here are three cheetsheet for accellerating data science workloads
A recap of Dr. Thomas Wiecki’s talk at PyData Berlin 2023 on how Bayesian statistical modeling can help solve marketing’s biggest problems, enabling efficient…
Pseudocode for common MCMC algorithms
A recap of the Microsoft Responsible AI Toolbox webinar, which introduced two new open-source tools Responsible AI Mitigations, a Python library for implementing…
A recap of the NLP.IL talk by Moran Beladev and Elina Frayerman on building a multilingual multi-label topic classification model for reviews data in Booking.com, using…
A recap of the NLP.IL talk by Nir Hutnik on validating NLP data and models, discussing common issues such as data and prediction drift, sample outliers, and error analysis…
fake data
using entropy to quantify uncertainty in probabilistic models
Understanding the challenges and strategies in multi-task learning
Robust Regression Techniques
A review and study guide of the paper ‘Simplifying Neural Networks by Soft Weight Sharing’ by Nowlan and Hinton, which introduces a method for reducing the complexity of…
Set Up M1 MacBooks for DS & ML
command line cheat sheet macos + zsh + git
A collection of hacks and tips for extracting content from PDF files, including converting PDFs to images using the pdf2image library in Python, and other techniques for…
How do humans make powerful generalizations from sparse data when learning about word meanings, unobserved properties, causal relationships, and many other aspects of the…
A meme is an idea, behavior, style, or usage that spreads from person to person within a culture. A meme bank would be a zoo for cataloging and breeding memes.
Getting more from your digital marketing agency, my five cents on picking your agency.
Attention for sensor fusion
Excel 2019 for Marketing Statistics in pandas
Language models and explainability
Marketing Research Models
Customer Lifetime Value (CLV) models and Buy Till You Drop (BTYD) models are used to estimate the total value a customer brings to a business over their entire relationship.…
Storytelling and other essentials,
Python Graph Cookbook
Stochastic Gradient Descent - The good parts
The WaveNet paper is kind of old. Yet it seems to come up in various contexts. Some thoughts on this.
Creating Citation Web Components
Wikipedia Hackathon notes
An algorithm for Inlining Citations for Wikipedia articles.
Transfer learning in NLP
writing better code = writing more readable code.
I’ve been excited about JSON-LD, when it came out as it offered a much cleaner way to do SEO. Only later did I realize that it is also a great format to worked with linked…
Different Multilevel Models Types
Concepts, slide commentaries and Lecture notes on Automatic text Summarization by Masa Nekic
Modeling Events.
How to blog like a life-hacker.
A review of the arXiv:2010.11967 paper by Chenguang Wang, Xiao Liu, Dawn Song
Review of the paper on dot product attention for the deeplearning.ai NLP specialization.
Just a rant at numpy and scipy breaking when I really needed them.
numpy
scipy
A collection of intuitions and insights about deep learning, including the functional form of feedforward neural networks with RELU activation, and how they can be…
A quick introduction to bash brace expansion, a powerful feature that allows you to generate multiple strings from a single pattern, making it easier to create lists of…
Just a little rant on Pandas various contexts
A quick tip on how to avoid cross site scripting (XSS) errors when using the Jupyter local runtime with Google Colab, by configuring the Jupyter notebook server to allow…
Post description
Inspired by Mixture of Experts (MoE), Lobeformation proposes boosting the Signal-to-Noise Ratio (SNR) by partitioning models into specialized sub-topologies (‘lobes’) that…
A/B testing cost and risks and some recommendation.
A recap of the talk on modern speech recognition, which covered the key principles of automatic speech recognition, the impact of self-supervised learning on the field, and…
This paper introduces a simple yet effective regularization technique called dropout, which helps prevent overfitting in deep neural networks by randomly dropping units…
In this paper the authors, investigated the effect of increasing the convolutional network depth on the accuracy in the large-scale image recognition setting. The authors…
A recap of the talk on efficient long-text understanding with short-text models, which covered the challenges of processing long sequences with transformer-based language…
How to model human cognition using probabilistic inference
A recap of the talk on SCROLLS, which covered the challenges of evaluating long-range reasoning in language models, the introduction of the SCROLLS benchmark for…
The introduction talk of the NLP IL meetup at Intuit, which provided an overview of the meetup’s goals, the speakers, and the topics that would be covered in the subsequent…
A detailed recap of the NLP IL meetup at Intuit, which included talks on long-range reasoning in language models, efficient long-text understanding with short-text models…
Effective time management is crucial for success in both personal and professional life. With the right approach, you can achieve more in less time while maintaining a…
A comprehensive review, study guide, and deep-dive into the SNA classic ‘Algebraic Models for Social Networks’ by Philippa Pattison, covering key concepts such as algebraic…
