Chelsea AI Ventures graph-of-skills-claude-plugin
0 stars MarkdownA Claude Code plugin for Graph-of-Skills. It connects the agent to the skill lookup service and adds one short skill that tells the agent to check for a written method before unfamiliar work.
AI Resources
Practical guides, articles and tools for your AI journey
We publish on the Chelsea AI newsletter, on what is working in production AI and what is not. It is where we named the Stochastic Man-Month Rule: adding AI to a workflow amplifies the verification workload faster than it speeds delivery. Articles, podcast appearances and recorded sessions are collected below, each with a summary so you can tell whether it is worth your time. Then the books our director has written, and the talks we have given.
Browse by topic
Four threads run through everything we publish. Each one collects the articles behind a part of what we do.
-
Strategy & ROI
Where AI investment pays back, and where it stalls.
18 articles
-
LLM & Agent Engineering
Retrieval, evaluation, agents and open models in production.
14 articles
-
Team & Capability
Human-in-the-loop working, skills and how teams adopt AI.
10 articles
-
Security & Governance
Attack surface, privacy and keeping AI systems accountable.
4 articles
Books by Our Team
Our training programs are built on practical implementation experience featured in bestselling books authored by our team
Retrieval Augmented Generation, The Seminal Papers
Published: March 2026 | New Release
Principles for architecting reliable and verifiable AI
Retrieval Augmented Generation (RAG) is a standard process for grounding LLM prompts in user-specified content rather than relying only on a model’s training data. RAG has grown from a simple prompt engineering workflow into a sophisticated set of data analysis, storage, and retrieval techniques. Retrieval Augmented Generation, The Seminal Papers explores foundational research papers that explain why RAG works, how it’s built, and what makes it different from other approaches.
Focus areas include:
Retrieval Augmented Generation AI Architecture Verifiable AI
Generative AI with LangChain (2nd edition)
Published: May 2025 | Amazon Bestseller in Programming
Build production ready LLM applications and advanced agents using Python and LangGraph
A practical guide to leveraging LangChain and LangGraph for GenAI implementation, with real-world examples ranging from customer support to data analysis. The 2025 edition features updated code examples and improved GitHub repository.
Focus areas include:
Enterprise-grade LLM application architecture Prompt engineering best practices RAG implementation for knowledge augmentation Custom agent development Production deployment strategies
Machine Learning for Time Series
Published: October 2021 | Industry Standard Reference
Forecast, predict, and detect anomalies with state-of-the-art machine learning methods
Use Python to forecast, predict, and detect anomalies with state-of-the-art machine learning methods. This comprehensive guide covers everything from data preprocessing to advanced models for time-dependent data. The included tutorials range from simple forecasting to complex deep learning architectures for time series analysis.
Focus areas include:
Anomaly detection systems Forecasting methodologies Feature engineering for time-series Deep learning approaches Production deployment patterns Time series preprocessing techniques LSTM and RNN architectures
Artificial Intelligence with Python Cookbook
Published: October 2020 | BookAuthority Best-Seller
Proven recipes for applying AI algorithms and deep learning techniques
Proven recipes for applying AI algorithms and deep learning techniques using TensorFlow and PyTorch. The practical cookbook approach provides ready-to-use solutions for common AI challenges, from computer vision to natural language processing, with complete code examples and detailed explanations of implementation considerations.
Focus areas include:
Deep learning fundamentals Computer vision applications NLP implementation techniques Reinforcement learning Model optimization strategies Transfer learning approaches Hyperparameter tuning
Software we publish
Everything here is public on GitHub, free to read and to use. The tools come out of consulting work. The rest is the code that goes with the books.
Tools
wiggum
2 stars ShellA self-driving agent loop for Claude Code. It turns an issue description into a workplan, implements it one task at a time, runs the verification commands the project already uses, fixes what fails, and commits as it goes.
langgraph-kirokuforms
1 star PythonHuman-in-the-loop for LangGraph. A workflow pauses, asks a person to fill in a form, and resumes with their answers, talking to KirokuForms over the Model Context Protocol.
Code from the books
rag-seminal-papers
7 stars Jupyter NotebookNotebooks working through the papers that retrieval augmented generation is built on, one implementation per paper.
Companion to Retrieval Augmented Generation: The Seminal Papers, Manning 2026
generative_ai_with_langchain
1.4k stars Jupyter NotebookCode for building LLM applications and agents with Python, LangChain and LangGraph, covering the examples in the book end to end.
Companion to Generative AI with LangChain, 2nd Edition, Packt 2025
Machine-Learning-for-Time-Series-with-Python-Second-Edition
3 stars Jupyter NotebookForecasting and time-series modelling in Python, updated with conformal prediction, causal inference, hierarchical models, and time-series foundation models.
Machine-Learning-for-Time-Series-with-Python
23 stars Jupyter NotebookForecasting and time-series modelling in Python, from classical methods through to deep learning.
Companion to Machine Learning for Time Series, Packt 2021
Artificial-Intelligence-with-Python-Cookbook
4 stars Jupyter NotebookRecipes across the breadth of applied AI in Python: supervised and unsupervised learning, probabilistic modelling, computer vision, language and speech.
Companion to Artificial Intelligence with Python Cookbook, Packt 2020
Speaking & Education
The GenAI Build Lab: Build Production-Ready RAG with Open Models
Packt Publishing, 2026
Focus: Production RAG on small open models: hybrid retrieval and reranking, corrective retrieval, RAGAS evaluation, and open guardrail models
Designing secure, customised, and sovereign LLM workflows
DataFest Tbilisi 2025
Focus: Workshop on cascaded agentic pipelines, grammar-enforced structured output, guardrails and golden-set evaluation on local CPU
See the event pageWhy open-source LLMs are the future of enterprise AI
DataFest Tbilisi 2025
Focus: Moving from public APIs to on-premise open-source models, and where small language models fit
See the event pageMoving Beyond Statistical Parrots: LLMs and Their Tooling
Put Generative AI to Work 2023, ODSC 2024, Data Science Week Amsterdam
Focus: Open-source LLMs and enterprise implementation
Future of Data Science and LLMs
PyData London 2023
Focus: Panel discussion on open-source AI technologies
Experimentation at Scale at loveholidays
RE•WORK, 2023
Focus: Running dozens of concurrent A/B tests on an in-house platform, at low bias and reduced lag
Watch the recordingMastering Machine Learning Techniques for Time-Series Analysis
PyData Bristol 2023
Focus: Feature extraction, forecasting and anomaly detection for time series in Python
View the slidesTime-Series Development
University of Oxford 2023
Focus: Guest lecture on the Artificial Intelligence: Cloud and Edge Implementations course
Strategic AI Implementation
Data & Analytics Summit EU 2022
Focus: Enterprise AI strategy and implementation roadmaps
Time-Series in Python: Preprocessing and ML
ODSC 2022
Focus: Machine learning techniques for time-series data
ML: The Good, The Bad and the Ugly
DataFest Tbilisi 2019
Focus: Where machine learning delivers in production and where it goes wrong
Podcast appearances:
- The agent autonomy trap (Agent Engineering, August 2026)
- What makes an ideal job candidate (Would You Data Scientist, January 2022)
- Time series prediction (Data Professor, January 2022)
Looking for a speaker for a conference, an internal session or a podcast? We take a few each year, on generative AI, MLOps and time series. Get in touch