I · Isolate
Set the reach of an identity.
Define which records, tools, and systems that identity may access.
The AI Architect Circle podcast
System Design with AI — Live with the Author
Bring a real design decision. Work through the reasoning with Rohit Bhardwaj. Join a free live conversation about system design, AI architecture, and decisions you can explain and defend.
Saturdays at 9:00 a.m. Eastern · Live on Zoom
With an enterprise AI architect, O’Reilly instructor, and author of System Design with AI Interview Guide. No book purchase or advance preparation required.
Inside the October 31 conversation
From an instruction to a boundary you can verify.
A prompt can describe the rule. The architecture needs a way to enforce it. We’ll begin exploring the next four LLM-SHIELD stages through practical design questions.
9–10 a.m. ET
The live podcast
Examine the boundaries, failure behavior, and evidence behind an AI system you can trust.
10–11 a.m. ET
Open questions with Rohit
Ask about system design interviews, preparing for AI roles, or a decision from your own project. Suggest a topic for a future episode.
Times are Eastern, anchored to America/New_York. Check Zoom for available dates and your local time.
E · Empower
Decide what may execute.
Make the authority for an agent’s actions explicit.
L · Live-Defend
Plan for a failed control.
Work through how the system should respond when a protection stops working.
D · Demonstrate
Bring evidence to the claim.
Identify the tests and observations that show a boundary holds.
Watch · Learn · Bring a question
AI Architect Circle on YouTube
Watch a practical lesson before the next conversation. Start with the Scope lesson from LLM-SHIELD, then bring a question about a boundary in your own system.
20:34LLM-SHIELD · Scope lesson
AI Agent Security: Why Prompts Aren’t Enough | LLM-SHIELD
Explore how to turn a business promise into a boundary that services enforce and tests can verify, with order-management and patient-record examples.
20 minutes · Opens on YouTube in a new tab.
Put the lesson to work
Read the companion design guide →Get the Agent Action Review PDF →The guide is open to everyone. The PDF requires a free account.
Keep learning between live sessions.
Subscribe to AI Architect Circle on YouTube for new lessons and published podcast videos.
Video thumbnails are served by YouTube. Watching a lesson opens YouTube in a new tab. Privacy details.
What to bring
A specific scenario gives us something useful to reason through.
An unfinished idea, an unclear requirement, or a difficult trade-off is welcome. You do not need a polished presentation.
“Where should I start?”
Practice framing the problem and explaining a first design choice before adding more detail.
“What am I trading away?”
Compare approaches and identify the constraint that makes one more appropriate.
“What should this change?”
Examine whether a capability helps the goal and which boundaries or evidence it requires.
Your host
Work through the reasoning with Rohit.
Rohit Bhardwaj is an enterprise AI architect, O’Reilly instructor, and author of System Design with AI Interview Guide.
The series is a space to make assumptions visible and connect technical ideas to practical decisions. Bring curiosity and a question you are willing to examine.
More about RohitA useful way to frame your question
“We are trying to achieve this outcome, under these constraints. Here is the option we are considering. The part I am unsure about is…”
A few details before you join
Do I need to be an architect already?
No. Software developers, architects, technical leaders, and professionals preparing for system design interviews are welcome. Bring a clear question and a willingness to examine the reasoning.
Do I need to read the book first?
No book purchase or advance reading is required. You can use a free guide or worksheet to prepare if it helps you frame a question.
How do I register?
Open the Zoom registration page, choose an available session, and complete the registration there. Zoom provides the joining instructions. Newsletter signup is a separate choice.
Will the session be recorded?
The session will be recorded and may be published as part of the AI Architect Circle podcast. Audience contributions may appear in the recording, so please avoid sharing confidential information. Published videos can be found on the YouTube channel.
What happens during the open Q&A?
For the October 31 session, the 9–10 a.m. podcast is followed by open Q&A from 10–11 a.m. Eastern. Bring interview questions, AI career preparation questions, or architecture decisions from your projects.
What does Eastern time mean for me?
The recurring schedule is anchored to America/New_York, including its daylight-saving changes. Use the date and time-zone selector on the registration page to check the local time for your session.
Before the conversation
Find a question to bring.
How to Frame a System Design Problem Before Choosing Components
Turn an ambiguous booking request into explicit outcomes, constraints, and a first decision. Includes a printable Design Decision Worksheet.
Your Retrieved Sources Disagree. What Should the System Do?
Separate relevance from applicability when retrieved policies conflict. Make authority, scope, effective dates, and uncertainty visible.
Define the Boundary Between an Agent Draft and a Real Action
Specify what an agent may prepare, what authorizes execution, and what happens when approval or verification is unavailable. Includes a printable review worksheet.
Continue the conversation
Hear about the next conversation.
You do not need to follow every announcement. Focus on the developments that change a decision, the evidence behind them, and the next question worth asking.
Read the published brief