Pillar 01
AI Software Engineering
Designing the software systems around models: architecture, evaluation, testing, deployment, observability, and product-facing reliability.
Carlos Daniel Jimenez
The Probability Engine
I write about AI Software Engineering: how MLOps evolves into LLMOps, how AI systems are designed and operated on GCP, how edge inference changes production constraints, and how NLP plus LLMs can be used to study music seriously.
Pillar 02
LLMOps and GCP
The operational transition from classical ML to LLM systems, with a practical bias toward evaluation, orchestration, and production architecture on GCP.
Pillar 03
Edge, Agentic AI, and Music
Two high-signal side lines: inference on constrained hardware and computational analysis of music with NLP and LLMs.
Reading map
How the site is organized now
The blog is intentionally narrower than before: AI Software Engineering is the center, with LLMOps, GCP, edge systems, agentic AI, and music analysis as connected lines of work.
LLMOps evolution
Where older MLOps assumptions break, and what replaces them in systems built around prompts, tools, traces, and evaluations.
AI Software Engineering
The engineering layer that turns models into dependable systems.
GCP, edge, and agentic systems
Architecture choices, inference constraints, and autonomous workflows once the cloud is only part of the story.
Music with NLP and LLMs
Computational analysis of lyrics, semantic structure, and the limits of language models as interpretive tools.
A technical breakdown of CPU, GPU, TPU, and Edge AI hardware tradeoffs for training and inference workloads — with real-world cost data and a deep dive into Raspberry Pi 5 + Hailo-10H.
A rigorous walkthrough of ISLP Chapter 2 fundamentals — from the formal definition of f(X) to the bias-variance decomposition, Bayes classifiers, and KNN — with Python code, real datasets, and connections to epistemology and learning theory.
Attention Windows: A Novel Framework for Measuring Narrative Cognitive Load in Beatles vs Pink Floyd
Abstract This research introduces Attention Windows, a novel framework for measuring the cognitive span required by listeners to follow lyrical narratives. How long can a theme …
Part 1: Philosophy, project architecture and orchestration with Hydra + MLflow. Steps for preprocessing, feature engineering, hyperparameter tuning and model registry.
Part 2: CI/CD with GitHub Actions, W&B vs MLflow comparison, complete containerization with Docker, and production-ready API architecture with FastAPI.
Part 3: Model selection strategies, advanced testing, production patterns, data drift, model monitoring, and production readiness checklist.
Editorial stance
What this blog is trying to do
I treat AI as a software and systems problem, not just a modeling problem.
I care about the shift from MLOps to LLMOps because operations change when prompts, tools, traces, and agents become part of the system.
I write about GCP and edge hardware because architecture matters as much as algorithms once cost, latency, and maintainability enter the picture.
I keep the music line of work because it pressure-tests what NLP and LLMs really understand when language becomes metaphor, rhythm, and structure.