Kuan Yu — Projects

Why we start doing this

Generalist AI Engineers are valuable now because they can move from customer problem to prototype to production. Over time, the field will likely fragment into sharper deployment, evals, data, LLMOps, and harness roles.

Software 3.0 reflection Natural language, code, and agent workflows are merging into a practical engineering discipline.
Claude Code workflow hacks The new leverage comes from designing the operating loop around frontier coding agents.
Claude Code for Entrepreneurs demand Operators and founders want people who can translate AI tools into usable business systems.
Claude Code API infrastructure Reliable model-powered workflows need runtime, monitoring, and deployment infrastructure.
Agent feedback loops The best AI systems improve through tight eval, feedback, and iteration loops.

AI engineering roles

AI Engineer Builds model-powered product workflows across frontend, backend, prompts, tools, and UX.

AI FDE Embeds with customers, scopes deployments, ships prototypes, and drives production adoption.

LLMOps Engineer Owns model runtime, observability, cost, reliability, routing, and release operations.

Evals Engineer Designs tests, datasets, scoring loops, and regression gates for model behavior.

AI Data Engineer Builds the ingestion, retrieval, transformation, and quality layer behind AI systems.

Harness Engineer Creates agent tools, command loops, scaffolds, and repeatable workflows around models.