Designing Effective Multi-Agent Architectures
Why multi-agent systems fail in production—and collaboration patterns that actually work beyond the prompting fallacy.
Read on SubstackNicole Koenigstein
AI researcher and practitioner in agentic AI — articles, books, teaching and speaking
Welcome to my personal website. I'm passionate about helping others to navigate AI Agent systems. I'm working across research, consulting, teaching, and direct system implementation to build reliable, production-ready AI systems. I focus on multi-agent architectures, evaluation, safety, and long-term system behavior.
I served as an external evaluator for a European Commission AI Grand Challenge and advised IOSCO on generative AI in regulated environments. I'm on advisory boards for leading AI and quantitative finance conferences. And regularly deliver invited talks and technical workshops across academia, industry, and international events.
I'm the author of Math for Machine Learning and Transformers in Action (Manning Publications), and Transformers: The Definitive Guide and AI Agents: The Definitive Guide (O'Reilly Media). Through my books, articles and teaching, I aim to make complex topics more accessible.
O'Reilly Radar
Articles on agentic AI and multi-agent systems, published on O'Reilly Radar · also on my Substack
The Hidden Cost of Agentic Failure
How probability compounds across agent handoffs, and why architectural debt shows up as instability and cost overruns at scale.
Read on SubstackBooks
Transformers in Action
A beginner-friendly guide to LLMs — summarization, classification, text generation, and responsible LLMs.
Transformers: The Definitive Guide
For intermediate to advanced readers — transformers in time series, RL, segmentation, and image/video generation.
AI Agents: The Definitive Guide
Design, deploy, and maintain production-ready LLM agents—architectures, tool integration, evaluation, and scaling beyond the demo stage.
Courses
Develop Self-Improving AI Agents with Reinforcement Learning
Hands-on live training on RL for LLM agents—rollouts, RULER rewards, GRPO/GSPO, and MCP tool use.
View on O'ReillyHands-On Multimodal AI
Build applications with text, image, audio, and video—LVLMs, transcription, and multimodal RAG.
View on O'ReillyAI Agents Bootcamp
Two-day bootcamp on designing and deploying enterprise agentic systems with LangGraph and multi-agent patterns.
View on O'ReillyLLMs in Finance Masterclass
Three-day live masterclass in Central London on applying LLMs in finance—International Faculty of Finance (IFF), June and October 2026.
View agendaMath for Machine Learning
Hands-on liveProject series covering vectors, matrices, calculus, and probability for ML.
EnrollAI in Finance Bootcamp
Intensive bootcamp on applying AI and machine learning in investment management and capital markets.
Learn moreLLMs in Finance
Learn how large language models and generative AI are transforming quantitative finance.
Learn moreAI Agents in Finance
Certificate program on designing and deploying agentic AI systems for finance workflows.
Learn moreGenerative AI and Agentic AI for Finance Certification
Certification cohort on generative and agentic AI applications in finance, with Packt Publishing.
View eventLatest posts
Building safer and smarter LLM agents with enhanced moderation pipline
In this blog, you'll explore how to design a robust and secure agent framework for interacting with LLMs and users. The code demonstrates integrating tools like Google Search, Wikipedia, and a calculator with advanced safety layers for moderation and compliance. This approach ensures the agent can reason, respond accurately, and adhere to ethical guidelines while preventing unsafe or malicious inputs and outputs. It integrates safeguards like content filtering, jailbreak detection, and contextual input moderation while maintaining the functionality of a dynamic ReAct agent.
Read moreMulit-agents in LangGraph for Investment Analysis
In this post you will learn how to do investment analysis with a multi-agent-setup. For this, you will be using the following tools and will also learn the following: Exa, after account login, get your API key here. To find the exact content you're looking for on the web using embeddings-based search. SerpApi here, after account login, get your API key to do look for existing patents. Python REPL, please note that Python REPL can execute arbitrary code on the host machine (e.g., delete files, make network requests). Use with caution. Tools to access and write to a .txt file and create a plot of historical prices. How to define utilities to help create the graph. How to create a team supervisor and the team of agents.
Read more