Julie Qiu: AI as a 'Thinking Partner' for Large-Scale Engineering Systems

Julie Qiu argued at QCon AI that AI can function as a practical "thinking partner" for teams managing large‑scale engineering systems, offering the cognitive "RAM" needed to synthesize sprawling legacy context and speed architectural choices. She framed the case around concrete tooling roles that let engineering leaders offload repetitive synthesis and exploratory work without losing oversight.

In her presentation Qiu named five distinct AI roles — Archaeologist, Experimenter, Critic, Author, and Reviewer — each mapped to a specific cognitive task. Archaeologist helps unearth and summarize legacy context; Experimenter pressure‑tests designs and hypotheses; Critic challenges assumptions and finds gaps; Author drafts proposals or code artifacts; and Reviewer inspects outputs for correctness and alignment. Together, these roles address the specific burden of understanding and evolving very large codebases.

Qiu said those roles are particularly useful for teams that must coordinate changes across hundreds of code repositories; she cited a target scale of more than 400 repositories as an example of the cognitive load these AI roles can help manage. The approach emphasizes augmenting human decision‑making — accelerating high‑level architectural decisions and surfacing tradeoffs — rather than automating ownership away from engineers.

Julie Qiu holds the title Uber Tech Lead for the Cloud Software Development Kit (SDK) at Google, where she builds client libraries and command‑line tools across language ecosystems to interact with Google Cloud products. She is also a Senior Staff Engineer at Google and previously served as a tech lead on the Go programming language team, where she led efforts on Go's vulnerability‑management support and the pkg.go.dev package discovery site.

The talk was part of QCon AI, a practitioner‑led conference focused on the engineering discipline required to scale AI workloads safely and on sharing operational playbooks and failure metrics used in production. The recorded presentation runs 42:10 and is available for download from the conference materials. For engineering leaders and platform teams, Qiu’s framework offers a practical path to reduce cognitive overhead: assign AI distinct, auditable roles to synthesize history, run experiments, critique designs, produce drafts, and review results — so humans retain final judgment while moving faster on architecture and system evolution.