Conference affiliation: Deployed Engineering Lead · Cognition · 2026
On this page
Jia Wu is a deployed engineering lead at Cognition who helps enterprises turn AI coding agents into measurable improvements in software delivery. Wu works across Devin and Windsurf, connecting autonomous coding systems to the testing, maintenance, migration, and organizational constraints that determine whether software actually ships.
From scientific computing to enterprise AI
Wu studied computational biochemistry at Western University and earned a master’s degree in mathematics at the University of Waterloo. Early research combined scientific software, data analysis, and machine learning, including neural-network approaches to identifying peptides in mass-spectrometry data.
After full-stack development at Eighty8 Ventures and machine-learning engineering at AUSPRE, where Wu worked on speech-processing systems and data pipelines, Wu became a solutions architect at Tyk Technologies. There, Wu helped enterprise customers evaluate API infrastructure and integrate existing authentication and business logic. Wu authored Tyk’s custom Go plugin repository, simplifying how customers build and bundle middleware for its API gateway, and advocated open standards, clear microservice boundaries, and internal platforms designed around business outcomes.
Wu subsequently worked in deployed engineering at Codeium and Windsurf before joining Cognition through its July 2025 acquisition of Windsurf. A public professional update describes recognition as Cognition’s leading customer-engineering revenue driver while crediting colleagues across sales, engineering, and leadership.
What makes coding agents useful inside enterprises
Forward-deployed engineering as a product feedback loop. Wu embeds with customers to identify high-value work, adapt agents to existing systems, and return recurring deployment problems to product teams. This separates customer-specific workarounds from missing capabilities that should shape the roadmap.
Software delivery beyond code generation. Wu treats producing code as only one component of enterprise engineering. Useful agents must also navigate existing codebases, write tests, support review and deployment, triage alerts, and modernize legacy systems, including environments built around COBOL and JCL.
Measurable outcomes over token consumption. Wu evaluates deployments through completed migrations, pull-request throughput, delivery timelines, and productive engineering capacity. High model usage alone does not establish that an organization shipped consequential work or improved its operations.
Human direction with autonomous execution. Wu combines interactive development with cloud agents that can carry out longer-running implementation and testing tasks. Cognition’s Devin integration with Windsurf illustrates how engineers can set priorities, inspect results, and delegate execution while retaining responsibility for quality.
