Approach on Tech Foundations

Combined Approach

Mon, 01 Jan 0001 00:00:00 +0000

Structured AI development approach:

Causal stack framework: the causal-stack set of notes under ai-engineering/evaluation/causal-stack, with Causal Stack Operating Model as the top-level map. It links language and system-level causal properties to fault families, evaluation methods, and operational controls, so it shows where the harness, platform, and methodology fit.
Evaluation harness and platform
Methodology and operating model
Observability and test pyramid
Experimentation and scientific foundation as well as Experimentation approach to delivery
Architecture and engineering principles
Cross-functional alignment with shared taxonomies and tools (frameworks, patterns, delivery)

and approach of design: Exploratory Design:

Cross Domain Mindset

Mon, 01 Jan 0001 00:00:00 +0000

Short answer: yes, they absolutely apply in ML, but some of them show up in slightly different ways and are even more important because of uncertainty around data and models.

Edd

Mon, 01 Jan 0001 00:00:00 +0000

resources

https://cookbook.openai.com/examples/partners/eval_driven_system_design/receipt_inspection

and also evaluation-driven system design / system delivery

Eval-Driven Development (EDD) is the architectural answer to the non-deterministic nature of AI.

Empirical Nature

Mon, 01 Jan 0001 00:00:00 +0000

AI Systems Require Behavioral Evidence

This note defines a working principle for designing, evaluating, and operating AI systems.

Short

An AI system’s behavior is implemented by its complete runtime composition, including model weights, prompts, code, data, tools, policies, and state.

Experimentation

Mon, 01 Jan 0001 00:00:00 +0000

Disciplined AI Experimentation

Objective: Define the systemic framework for transitioning AI development workflows from uncontrolled, intuition-based iterations to evidence-driven, decision-grade engineering.

Experiments Vs Features

Mon, 01 Jan 0001 00:00:00 +0000

Resources

Input

downloaded video “Stop Managing AI Projects Like Traditional Software” and https://www.youtube.com/watch?v=R_HnI9oTv3c

https://www.youtube.com/watch?v=k98gDjYbSaU Failure is a funnel, also downloaded

Workflow Design

Mon, 01 Jan 0001 00:00:00 +0000

Designing workflows:

Interaction-point design approach for AI workflows. The idea was to expose uncertain pipeline decisions – retrieval candidates, context assembly, validation, review judgments, and failure labels – as inspectable artifacts before automating them. That gave us a way to compare variants, capture domain feedback as traces, and turn stable patterns into evaluation criteria, operational controls, or automated policies.