Architecting Intelligence
Articles about Architecting Intelligence
Enterprise AI Model Strategy
Why the enterprise question moved from which model to which model for which work, and how to price the off-switch.
Scaling Performance Across 100,000 GPUs
Bringing 100,000 GPUs Into Production: A Scalable Acceptance Architecture
Building Depth #3 — The Evolution of Re-ranking and Slate Optimization
Cracking ML System Design Interviews · Deep dives into Part 2: Design a Search and Ranking System
Architecting Reinforcement Learning for LLMs — Part 3: Reward Modeling: Training, Scaling, and Failure Modes
A Note on the Series and This PartArchitecting Reinforcement Learning for LLMs is a ten-part, research-driven series exploring the algorithms, feedback systems, training pipelines, and architectures used to align language models, improve reasoning, and train agents.
Architecting Foundation Models: An Architecture Map
Introducing a 14-part series of live architecture talks and detailed Substack articles
Looped Transformers: Can Models Think Longer Without Generating More Tokens?
Can models think longer without generating more tokens? In the first episode of Architecting Foundation Models, we explore looped transformers: how they reuse layers for iterative computation, how recurrent depth differs from a standard transformer, and what it could mean for reasoning and test-time compute. We also discuss ideas from Universal Transformers and recent research on latent reasoning.
Architecting Agentic AI — Part 1: The Agentic Design Framework
A Note on the SeriesArchitecting Agentic AI studies one system through six connected views: product experience, agent intelligence, software runtime, data and memory, security and evaluation, and production economics. Part 0 introduced those views and the eleven-part roadmap. In this part, we establish a decision framework that the remaining articles will use. A single reference implementation will grow through the series; today we define its first task contract and a minimal bounded loop.
Building the Architecting Intelligence Learning Ecosystem
Premium research, role-based learning, realistic interview practice, online classes, and mentorship—connected in one journey.
Building Depth #1 — The Evolution of Retrieval Systems in Search & Recommendation
A Note on the Series and Parts:Cracking ML System Design Interviews is a practitioner-driven series focused on production-scale ML and AI systems for Staff- and Principal-level interviews.
We Built Agents That Can Attack. Now We Need Agents That Can Defend
Where agent safeguards break down, why deliberate misuse changes the problem, and how to architect a defensive AI system