Jauzo - Mikko Ahonen – Honest, pragmatic & innovative IT professional
Is working with LLMs more like leadership or more like management ? The words management and leadership are used to convey many kinds of meaning, so it is important to elaborate carefully what we mean when we say that “working with LLMs is more like <x>”. For me, one of the most important aspects of leadership is having a long-term vision, and being able to communicate it. In other words, it…
The concept of eventual consistency often comes up when modern enterprise architectures are discussed, especially in the context of microservices. In typical configuration, an eventually consistent system is actually never guaranteed to be in consistent state. Therefore, it is more precise to say that the system is perpetually inconsisent, and at the minimum, saying that a system is eventually…
Complexity arises from having many tangled parts. Complexity means that you cannot realistically look at each part separately, because it can only be understood in relation with the other parts. Complexity may be intrinsic – directly related to the problem we are trying to solve. Or it may be incidental – related to how we are solving it. When we put these two types of complexity on a matrix, we…
We often talk about LLMs in terms of reasoning and generation. I have found it more helpful to think in terms of transformations and semantic layers. There appears to be some general principles that I have not found explicitly expressed elsewhere, on the relationship between semantic layers and transformations. Understanding them helps us to build more reliable systems with the help of LLMs. The…
Recently, a moderately high volume Django site started to fail on MultipleObjectsReturned. The offending model was an advertisement impression counter. The counters were created dynamically using get_or_create(). It was failing, because get_or_create() may create duplicates if the model does not have a unique constraints for the fields used in the filtering. And I did not have such a contraint.…
This post is eleventh part in a series about building and adopting a modern streaming data integration platform. In this post I will discuss how to think about development and maintenance costs in various parts of the architecture. Part 1: Why to Build a Data Integration Platform? Part 2: About Logging Part 3: Data Integration Methods Part 4: Role of Data Modeling in Data Integrations Part 5: Data…
This post is tenth part in a series about building and adopting a modern streaming data integration platform. In this post I will discuss how to release the master data locked in legacy systems, such as mainframes. Part 1: Why to Build a Data Integration Platform? Part 2: About Logging Part 3: Data Integration Methods Part 4: Role of Data Modeling in Data Integrations Part 5: Data Integration…
This post is ninth part in a series about building and adopting a modern streaming data integration platform. In this post I will discuss three important aspects of robust data integrations when using the logging metaphor: global ordering of messages, multi-message transactionality, and exactly-once semantics. Focus is on Apache Kafka, which is a common platform for implementing streaming data…
This post is eigth part in a series about building and adopting a modern streaming data integration platform. In this post I discuss the various communication patterns in real-time data integrations. Part 1: Why to Build a Data Integration Platform? Part 2: About Logging Part 3: Data Integration Methods Part 4: Role of Data Modeling in Data Integrations Part 5: Data Integration Architectures Part…
This post is seventh part in a series about building and adopting a modern streaming data integration platform. In this post I discuss our lost integrity on data integrity. Part 1: Why to Build a Data Integration Platform? Part 2: About Logging Part 3: Data Integration Methods Part 4: Role of Data Modeling in Data Integrations Part 5: Data Integration Architectures Part 6: Lightweight Reference…