The AI Practitioner Podcast

Prefer reading instead? The full article is available here. The podcast is also available on Spotify and Apple Podcasts. Subscribe to keep up with the latest drops.Large language models are powerful, but relying on prompts alone quickly becomes fragile and difficult to scale. As teams try to operationalize LLMs in real workflows, traditional documentation and ad-hoc prompting start to break down.In this episode, we explore a new paradigm introduced with Claude Skills: packaging workflows, instructions, and resources into reusable capabilities that LLMs can execute.You’ll learn:* Why traditional documentation is poorly suited for LLMs and why workflow-first instructions are more effective.* How Claude Skills structure tasks using a concise SKILL.md file that points to supporting files and scripts loaded on demand.* How teams can design and deploy skills to turn LLMs into reliable task executors rather than prompt-driven tools.By the end, you’ll understand how skills move us from prompt engineering to designing AI-native workflows.If you’d rather read than listen, the full article (with diagrams, code examples, and implementation details) is available on Substack:👉 Enjoyed this episode? Subscribe to The AI Practitioner to get future articles and podcasts delivered straight to your inbox. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit aipractitioner.substack.com

Prefer reading instead? The full article is available here. The podcast is also available on Spotify and Apple Podcasts. Subscribe to keep up with the latest drops.AI assistants are quietly reshaping how people discover products and documentation online. But most analytics systems treat AI bot traffic as noise, filtering it out instead of learning from it. In this episode/article, we explore how to uncover real user intent hidden inside AI assistant traffic and turn bot logs into actionable insights for product and SEO teams.You’ll learn:* Why AI assistant traffic is fundamentally different from traditional bot traffic, and why filtering it out creates a major blind spot in modern analytics* How prompts sent to tools like ChatGPT, Claude, or Perplexity translate into bot visits, and what these patterns reveal about real user questions, product research, and integration needs* A practical framework for analyzing AI bot logs, helping teams extract user intent signals that can inform documentation improvements, product decisions, and SEO strategyIf you’d rather read than listen, the full article (with diagrams, code examples, and implementation details) is available on Substack:👉 Enjoyed this episode? Subscribe to The AI Practitioner to get future articles and podcasts delivered straight to your inbox. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit aipractitioner.substack.com

Prefer reading instead? The full article is available here. The podcast is also available on Spotify and Apple Podcasts. Subscribe to keep up with the latest drops.Most agent systems reason well in the moment but fail to improve over time because they forget everything once execution ends. In this episode, we explore how to design long-term memory for LangGraph agents, moving beyond short-term context toward durable, structured memory that remains transparent and controllable. You’ll learn:* Why long-term memory is an architectural problem, not a prompt-engineering trick, and how different memory types (working, semantic, episodic, procedural) interact in agent systems* What LangGraph provides out of the box for memory management—and where it stops, especially when building agents that must persist, update, and reason over memory across sessions* How to implement schema-driven long-term memory with Trustcall, enabling safe extraction, controlled updates, and debuggable memory writes inside LangGraph nodesIf you’d rather read than listen, the full article (with diagrams, code examples, and implementation details) is available on Substack:👉 Enjoyed this episode? Subscribe to The AI Practitioner to get future articles and podcasts delivered straight to your inbox. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit aipractitioner.substack.com

Prefer reading instead? The full article is available here. The podcast is also available on Spotify and Apple Podcasts. Subscribe to keep up with the latest drops.Agent systems break down when simple workflows evolve into tangled 30+ node graphs with unclear dependencies and sequential bottlenecks. In this episode, we explore how to scale LangGraph architectures through strategic parallelization, modular subgraphs, and dynamic task distribution. You’ll learn:* When to use parallel execution vs. sequential flows and how to manage concurrent state updates with reducers?* How to structure multi-agent systems using subgraphs with either shared or isolated states?* When dynamic map-reduce patterns outperform static parallelization for variable workloadsIf you’d rather read than listen, the full article (with diagrams, code examples, and implementation details) is available on Substack:👉 Enjoyed this episode? Subscribe to The AI Practitioner to get future articles and podcasts delivered straight to your inbox. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit aipractitioner.substack.com

Prefer reading instead? The full article is available here. The podcast is also available on Spotify and Apple Podcasts. Subscribe to keep up with the latest drops.Autonomous agents promise efficiency, but without visibility and control, they risk costly mistakes before anyone can intervene. In this episode, we explore how to transform AI agents from opaque black boxes into steerable, inspectable systems using LangGraph’s human-in-the-loop capabilities. You’ll learn:* How streaming exposes an agent’s reasoning in real-time, from token generation to state transitions, building trust through transparency* How breakpoints enable surgical intervention at critical decision points, allowing humans to approve, reject, or correct actions mid-execution* How time travel lets you rewind to any prior state, fork alternative reasoning paths, and explore “what-if” scenarios without restarting from scratchIf you’d rather read than listen, the full article (with diagrams, code examples, and implementation details) is available on Substack:👉 Enjoyed this episode? Subscribe to The AI Practitioner to get future articles and podcasts delivered straight to your inbox. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit aipractitioner.substack.com

Prefer reading instead? The full article is available here.Demo agents are easy to build, until they crash mid-execution, lose conversation context, or explode your token budget. In this episode, we explore the three critical mechanisms that transform fragile prototypes into production-grade AI systems. You’ll learn:* How to optimize state with reducers and caching — managing growing state efficiently through composable update functions and skipping expensive recomputation through intelligent caching strategies.* How to implement persistence and memory — maintaining state across sessions, preserving conversation history, and ensuring agents remember what they’ve already done to avoid redundant work* How to build fault-tolerant systems with checkpointers — saving state at every step, resuming execution from any point, and recovering gracefully from failures without losing progressIf you’d rather read than listen, the full article (with code examples, implementation patterns, and debugging strategies) is available on Substack:👉 Enjoyed this episode? Subscribe to The AI Practitioner to get future articles and podcasts delivered straight to your inbox. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit aipractitioner.substack.com

Prefer reading instead? The full article is available here. The podcast is also available on Spotify and Apple Podcasts. Subscribe to keep up with the latest drops.Linear AI chains fail the moment reality gets messy: when APIs break, reasoning loops infinitely, or context is lost between steps. In this episode, we dive into how LangGraph reimagines agent design with stateful, graph-based reasoning that mirrors how scientists actually think. You’ll learn:* Why linear chains can’t handle non-linear thought or adaptive reasoning* How graph-based agents recover from failures using state, loops, and conditional logic* How LangGraph Studio and LangSmith provide full observability—from local debugging to production monitoringIf you’d rather read than listen, the full article (with diagrams, code examples, and implementation details) is available on Substack:👉 Enjoyed this episode? Subscribe to The AI Practitioner to get future articles and podcasts delivered straight to your inbox. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit aipractitioner.substack.com

Prefer reading instead? The full article is available here. The podcast is also available on Spotify and Apple Podcasts. Subscribe to keep up with the latest drops.Real-world AI agents fail differently than traditional software, silently, with confident hallucinations instead of error codes. In this episode, we explore how AgentOps adapts DevOps principles to handle the unique challenges of LLM-powered systems. You’ll learn:* Why agent systems require fundamentally different operations than traditional ML models* How the AgentOps lifecycle handles probabilistic reasoning and semantic failures* How to implement production-grade observability using MLflow’s tracing, prompt management, and evaluation capabilitiesIf you’d rather read than listen, the full article (with code, implementation details, and comprehensive examples) is available on Substack:👉 Like this kind of content? Subscribe to get future articles and episodes delivered straight to your inbox as soon as they’re published. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit aipractitioner.substack.com

Prefer reading instead? The full article is available here. The podcast is also available on Spotify and Apple Podcasts. Subscribe to keep up with the latest drops.Real-world decisions often involve conflicting objectives, like boosting sales while avoiding overstock. In this episode, we explore how multi-objective optimization (using NSGA-II with the pymoo library) can model and solve such dilemmas. You'll learn:* Why traditional dispatch strategies fall short* How Pareto fronts reveal optimal trade-offs* How to apply ASF and Pseudo-Weights to guide final decisionsIf you’d rather read than listen, the full article (with code, charts, and detailed examples) is available on Substack :👉 Like this kind of content? Subscribe to get future articles and episodes delivered straight to your inbox as soon as they’re published. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit aipractitioner.substack.com