In this article I explore Edge architectures utilizing Federated Learning and model delta transfer to acknowledge typical edge connectivity and badnwidth issues.
Federated Learning & Model delta transfer
Self-adjusting AIOps Solutions
Here’s a quick update on the LFEdge AIOps for Distributed Environments project:
https://medium.com/@andreas.spanner/towards-self-adjusting-aiops-systems-an-update-33e523d9686a
Headless Retail
Based on the discussions around the NRF – The Retail Big Show in Singapore this latest blog describes the future of Retail being Headless Retail and links also my two earlier blog posts to give the reader a cohesive and comprehensive view across technology, humans and the resulting user (UX) or better AI Agent Experience (AAX).
Based on the conversations during NRF and dedicated Retail dinners with startups and partners this is not a comprehensive list – there are more things coming, but Headless Retail is right in the center of it.
Omni-Channel is so Yesterday – Check out Agentic Retail
In this demo and associated blog post, I show how agentic Retail will change our shopping experience.
It covers things like:
- Autonomous negotiations and purchasing
- Comparing different shop and buyers agents’ performance
- Have you agent rendering product choices the way you like it best
- An evaluation that tests all permutations of shop and buyers agent, allowing to compare performances across:
- Store negotiation strategies eg relationship, competitive or premium
- Buyers personalities
- Buyers profiles
- Floor price
- Stock levels
- Margin pressures
- and more technical things like context engineering, RAG, time-to-first-token (TTFT), etc.
All experiments are tracked via MLFlow.
Please reach out and let me know what you think! Your feedback is a much appreciated gift.
Benchmarking LLM Agents for AIOps – The Chaos Engineering Way
In this article I am describing an Chaos Engineering way to test AIOps agents via an evaluation harnesses:
Consolidated Telemetry for AIOps
In this blog post I cover the current open source available storage backends for data ingestion and retrieval to support organisations toward AIOps for inference and training across anomaly detection, signal correlation, root-cause analysis and remediation as part of a codified bechmark harness under LFEdge/InfiniEdgeAI/AIOps: https://github.com/lfedgeai/AIOps
I am examing data ingestion performance
and data query performance:
Federated Learning (AI) for the Enterprise
In this article at redhat.com Yan Meng and I go through how kubernetes CRDs of the open source project Open Cluster Management (OCM) help to create, deploy and manage distributed and federated learning (AI) environments:
Let me know which FL use cases you are currently working on or would like to explore!
Thanks!
Andreas
OpenShift with GPU support setup with YAKKO
In this article I cover some pitfalls you need to be aware of if you want to ‘free’ your NVIDIA (or any other really) GPU from your HDMI port, because you want to use it for AI model work:
https://medium.com/@andreas.spanner/openshift-with-gpu-support-on-your-laptop-6552c3932b34
Geospatial use cases with Prithvi AI Models
In this post I showcase in collaboration with Michael Cawte (Principal at Section6) how to utilize geospatial AI capabilities based on the Granite Prithvi Model family.
Use cases based on IBM use cases are:
- Flood Prediction
- Earth Surface temperature prediction
- Canopy height prediction
https://medium.com/@andreas.spanner/geospatial-granite-models-on-openshift-4119e7c1b04d
All this is deployed on Red Hat OpenShift and was developed as a demo for a Fire & Emergency department.
Agile, DevOps & Enterprise Architecture – War or Peace?
I recently caught up with John Willis (co-author of the DevOps handbook). We were both on a call with a client and after that hosting the CNCF & DevOps Meetup in Wellington with our good friend BMK. On that day the good old question “Do organisations need Enterprise Architecture (EA) in times of Agile & DevOps?” came up again.
The timing for this question was perfect, because I spent many years in Enterprise Architecture and Agile, while Johns background is obviously DevOps and Agile, but we both have the same view: Yes, absolutely! And here’s why.
What is Enterprise Architecture?
This question is fundamental, because once you understand what Enterprise Architecture is you can derive the answer. The problem with the original question is really homemade.
Because many organisations call some of their architects Enterprise Architects, when they are really not, causes confusion. If you are looking after an Enterprise wide virtualisation platform for example you are NOT an Enterprise Architect in an Enteprise Architecture sense. I’ve written a more detailed piece on the different types of architects here.
Simply put, an Enterprise Architect helps define mission, vision, goals and the strategies to accomplish those goals. Mr & Mrs EA then help determine which capabilities (people, process & technology) the organisation needs to build in order to be able to execute the strategies that help accomplish the goals. A great framework depicting how this fits together is the means-to-end framework:
The ‘Tactical’ layer defines the programs, projects, products and services (i.e. initiatives) that build or uplift the desired capabilities. You can of course also find more scientific definitions of EA on the internet, but this is a simple and practical explanation of what EA is which works well for me.
And now I am handing over to Agile and DevOps
And that’s the moment where Agile and DevOps comes into play.
DevOps principles and the Agile way of working are methodology options (just like manual & waterfall) how you prioritise, define, build, test and deliver capabilities or capability uplifts. You could call Agile and DevOps also ‘enabling capabilities’ if you like.
And that’s really it. Simples.
Now, it’s easy to see how things get confusing when people who write Java Enterprise Edition (J2EE) based software get referred to as Enterprise Architects. But the problem is not with Enterprise Architecture, the problem is that those people are simply not Enterprise Architects.
If such a confusion exists in your organisation or with your customers, it might be worth running a workshop and getting the categorisation and scope of all the architects in your organisation examined and clarified. Examples of different types of architects are: systems, software, technical, infrastructure, operations, presales, delivery, business and enterprise architects.
Get them in a room to land on a common definition and understanding. It might get heated, but that’s OK. Terminally nice is way worse than passionate, respectful exchanges.
Keen to get your thoughts,
Andreas