Eric Hou is a software engineer focused on making AI coding agents effective collaborators in complex production systems. At AI Engineer World’s Fair 2025, he represented Augment Code, where his work centered on agent-assisted development, shared engineering knowledge, and the human oversight required to ship reliable software.
Before Augment, Hou engineered products at Applied Intuition, worked on automotive and mobility standards, and conducted computer-vision research. In 2018, he and Griffin Haskins developed Autocomm I/O, an edge-computing platform designed to run connected-device applications closer to their hardware. Those experiences exposed him to sprawling systems whose architecture and institutional knowledge exceed any individual engineer’s grasp.
How Hou approaches agentic engineering
- Mentoring the machine: Hou treats coding agents like junior engineers who lack organizational context: give them clear outcomes, architectural boundaries, relevant knowledge, and human review. His AI Engineer World’s Fair presentation illustrates the model through design-system work, incident response, RFC preparation, and a gRPC upgrade spanning multiple services.
- Parallel agent orchestration: Instead of forcing engineers to abandon feature work during an incident, Hou assigns separate agents to inspect logs, locate regressions with Git bisect, draft documentation, and investigate remediation. Engineers coordinate the work, evaluate proposed changes, and remain responsible for production readiness.
- Tasklist as a shared planning surface: Hou uses Augment’s Tasklist workflow to convert open-ended requests into editable execution plans. Engineers can inspect dependencies, revise tasks during implementation, and preserve reusable workflows while recognizing that sensitive operations and poorly defined investigations require closer supervision.
- Organizational knowledge infrastructure: Codebases, architectural decisions, documentation, issue trackers, and shared agent memories give automated systems the context needed to follow established engineering practices. Hou also advocates using inexpensive agent-built prototypes to compare competing approaches before committing to production architecture.
