Testing & agents / In progress
What years as an SDET taught me about building agents
The habits that make failures reproducible and success measurable matter even more when software chooses its own next step
Read the outlineENGINEERING / AUTONOMOUS SYSTEMS
Autonomous systems
in practice
Architecture, evaluation, and the decisions behind reliable agents
AWS & ML / In draft
What working in AI/ML at AWS changed about my engineering instincts
Model behavior is one part of the system — deployment, observability, interfaces, and failure handling determine whether the capability is useful
Read the outlineGames & engineering / Coming soon
Starting my career in games
Early work in games offers a particular education in state, performance, and users who take paths the design never anticipated
Read the outlineTesting & agents / Research draft
An agent that passes once hasn’t passed
A successful demonstration and a reliable system provide different kinds of evidence
Read the draftSystems & state / In draft
The hardest bugs live between steps
A sequence of reasonable actions can still produce an unreasonable result — the transitions deserve their own tests
Read the outlineCareer & practice / Coming soon
What I carried from games to testing to AI
Three stages of an engineering career, connected by the work of understanding systems when they leave the expected path
Read the outlineAWS & ML / Research draft
What belongs in an ML integration test?
A practical boundary between checking the model, checking the service, and checking the application that depends on both
Read the draftAWS & ML / In progress
Debugging systems with several plausible explanations
A bad result might come from the model, the context, the orchestration, or the infrastructure — debugging starts by separating those possibilities
Read the outlineAWS & ML / Research draft
What a production metric can’t tell you
Aggregate results can look healthy while a small but important class of failures remains invisible
Read the draftGames & engineering / In draft
Players find the paths you didn’t design
Unexpected behavior is useful evidence about the assumptions built into a system
Read the outlineGames & engineering / In progress
State is where the trouble starts
State that is easy to create can be difficult to inspect, restore, or reason about
Read the outlineGames & engineering / Coming soon
What I misunderstood about a career in games
A candid account of the distance between wanting to work on games and understanding the day-to-day engineering work
Read the outlineTesting & agents / Research draft
The test oracle problem, again
Before asking whether an agent passed, decide what evidence would make an answer correct
Read the draftTesting & agents / In progress
A useful failure needs a replay
The value of a failure depends partly on whether another engineer can investigate it
Read the outlineTesting & agents / In draft
Testing the decision to stop
Completion, escalation, and unnecessary continued action are distinct behaviors that need distinct evaluations
Read the outlineTesting & agents / Coming soon
From regression suites to evaluation sets
Some testing habits transfer directly to AI development — others need a different definition of coverage
Read the outline16 essays in development · Dates refer to drafts, not publication
02 / EXPERIMENTS
Systems under test
Working demonstrations of agent behavior, decision-making, and evaluation
Experiments will appear here as they are ready to share
IN DEVELOPMENT03 / ABOUT
Anthony DiRaddo
I build autonomous AI systems, with a focus on the engineering decisions that make them useful and dependable
My work sits at the frontier of AI software development: designing systems that act, use feedback, and execute toward a goal
LinkedIn