The IBM Z Deep Dive Series is a monthly, practitioner-focused technical webinar series designed to help professionals deepen their expertise in IBM Z technologies. Each session focuses on a specific technology topic and brings together subject matter experts through engaging and interactive discussions.
I recently attended the z/OS Deep Dive, the very first edition of this brand-new series, and wanted to share my key takeaways, insights, and how I plan to apply what I learned.
Why I Joined This Session
I joined this Deep Dive session because staying current with the z/OS ecosystem is essential to my role. As a Senior Technical Expert on IBM Z with over 25 years of experience, and as a five-time IBM Champion, I know firsthand that the platform evolves faster than many people realize. The topic aligned closely with my current work and professional interests, especially around z/OS simplification and the growing openness of the platform to industry-standard tools and practices.
I was also interested in hearing real-world examples, connecting with peers, and discovering new ideas that I can apply in my day-to-day role. Managing z/OS environments for an ISV that serves the banking industry means I need to anticipate where the platform is heading, not just react to changes after they happen.
I watched all six sessions on replay, which allowed me to go through each topic at my own pace and take detailed notes. The replay format worked perfectly for a full-day event of this depth.
Key Highlights
Highlight #1: IBM Project Polaris, A Paradigm Shift for z/OS Configuration
The session on Project Polaris, presented by Fiona King and Rolando Perez, was the most forward-looking of the day. IBM is reimagining how system programmers interact with z/OS by replacing traditional Parmlib definitions with industry-standard YAML files, managed through VS Code and Git, and guided by WatsonX Assistant for Z.
The live demo was compelling: a relatively new system programmer ("Chris") was able to complete an SMF configuration update, a task that would normally take four to five hours, in a fraction of the time, guided step by step by the AI assistant. The assistant didn't make changes automatically, it provided contextual guidance, opened the right file at the right location in the IDE, and helped verify the result. This is exactly the kind of experience that will make z/OS more accessible to early-tenure professionals without compromising the depth and rigor that experienced practitioners expect.
Highlight #2: IBM SMF Explorer, Modern Data Access for Everyone
The session by Moritz Mülhstädt on DataGatherer REST Services and IBM SMF Explorer was a standout for me. SMF data is the lifeblood of z/OS performance analysis, but working with it has traditionally required deep knowledge of native interfaces, assembler, and hierarchical data layouts.
With SMF Explorer, a Python library, anyone can now retrieve and process SMF data in a tabular format using just a few lines of Python code. Combined with Jupyter Notebooks, this opens the door to interactive data exploration, visualization with libraries like Plotly and Pandas, and even machine learning workflows, all without needing to write a single line of assembler or JCL. The REST API layer underneath means you can also build custom tooling or integrate SMF data into your existing observability stack.
Highlight #3: AI-Infused z/OS Is Real, And It's Free
The session by Anastasiia Didkovska and Steve Partlowon AI-Infused z/OS was an eye-opener. The AI Framework for z/OS, including a subset of Machine Learning for z/OS (Core Edition) and Common Data Provider, comes at no additional charge with z/OS. It enables two production-ready use cases today: AI-powered WLM batch initiator management (up to 70% queue time reduction) and AI-powered network outbound packet batching (up to 18% reduction in network interface calls).
What impressed me most is the trust-building approach: users can run AI in simulation mode to observe what AI would recommend before actually enabling it. The AI doesn't take over. It enhances, and you remain in the driver's seat.
Technical Deep Dive
Here are three notable technical concepts that stood out across three of the six sessions:
Project Polaris and Configuration as Code with YAML. Polaris converts all 70+ Parmlib definitions into a clean set of YAML files, enabling syntax checking, auto-completion in VS Code, and Git-based version control with approval workflows. Parmlib doesn't go away, but YAML becomes the primary interface. The integration with WatsonX Assistant for Z adds contextual, on-system guidance that draws from internal documentation rather than external sources, minimizing hallucination risks.
SMF Explorer and the DataGatherer REST API. The DataGatherer REST Services run as a separate z/OSMF server instance and expose SMF data as JSON via OpenAPI-compliant endpoints. SMF Explorer sits on top as a Python library that flattens hierarchical SMF records into Pandas DataFrames, with virtual fields for easier analysis. A public GitHub repository provides tutorial Jupyter Notebooks to get started.
AI Framework for z/OS. The framework is composed of five building blocks: Common Data Provider for data collection from SMF buffers, EasyNoSQL as a VSAM-RLS backed JSON database, ML for z/OS Core Edition for model training and inferencing on zIIP, an AI-based component running on each system in the Sysplex, and the z/OSMF AI Control Interface plugin. All training and inferencing run natively on z/OS, primarily on zIIP processors.
What I Plan to Apply
One idea that really stood out to me during the session was how SMF Explorer can democratize access to z/OS performance data. I especially enjoyed learning about the Jupyter Notebook tutorials available in the public GitHub repository, which provide a hands-on way to get started with real SMF data analysis using Python.
Being able to build lightweight, reusable Python notebooks for capacity planning, performance troubleshooting, or even training purposes is something I want to explore concretely. SMF Explorer could also complement our existing monitoring approach by making it easier for team members who are more comfortable with Python than with ISPF to contribute to system analysis.
After attending this Deep Dive, I would also like to explore Project Polaris as it enters its tech preview phase this summer. The shift from Parmlib to YAML, combined with Git-based workflows and AI-guided configuration, aligns perfectly with the direction we've been taking in our own z/OS environments, where we already use VS Code, SSH, Zowe, and Ansible. Polaris feels like the natural next step in that modernization journey.
Community Experience
One of the most valuable parts of the IBM Z Deep Dive Series is the opportunity to connect with others in the community and learn from different perspectives.
Being the very first edition of the series, there was a genuine sense of excitement around the event. The quality of the Q&A across all sessions was remarkable. Attendees asked sharp, practical questions about Sysplex requirements for AI, training data volumes, Parmlib coexistence with YAML, and whether AI could overtake system decisions (spoiler: it can't, guardrails are in place). Several of these questions sparked deeper reflections that I continued exploring on my own after the event.
As an IBM Champion, I see this Deep Dive Series as an excellent resource not only for my own continuous learning, but also as content I can share with my broader network: colleagues, clients, and the French IBM Z community. The replay format makes it easy to recommend specific sessions to people based on their role and interests.
At the end of the series, I earned my IBM Z Deep Dive badge, which helped recognize the learning experience and technical knowledge gained during the event.
Looking Ahead
I'm looking forward to the next IBM Z Deep Dive session on IMS, scheduled for June 17, 2026, and continuing to learn alongside the IBM Z community.
If you haven't registered yet, I highly encourage you to do so. Whether you attend live or watch the replays, the depth and quality of these sessions make them well worth your time. Register here : https://ibm.biz/ZDeepDive
See you at the next one!