Lei Zhang is Bloomberg’s head of technology infrastructure, overseeing the systems that support more than 9,000 engineers across developer productivity, reliability, telemetry, incident response, connectivity, and global infrastructure. His central challenge is turning faster AI-generated code into measurable improvements in software quality and delivery.
Zhang studied computer science at Stony Brook University and worked at Siemens Corporate Research before joining Bloomberg, where he progressed through derivatives-data management and developer-experience leadership into organization-wide infrastructure responsibilities.
Engineering ideas that define his approach
- Measure the entire delivery pipeline. Coding assistants accelerate prototypes, tests, and disposable scripts, but generating more code also increases pull-request queues and time to merge. Zhang measures productivity against review capacity, verification, and the ability to ship dependable changes.
- Use uplift agents to tackle neglected maintenance. These agents scan established codebases, propose patches, support migrations, and extend automated refactoring beyond pattern-based tools. Their usefulness depends on deterministic verification through tests, linters, and other reliable checks.
- Give incident-response agents real operational context. Effective troubleshooting requires access to logs, metrics, traces, feature flags, service dependencies, network topology, and service-level objectives. Zhang’s approach uses MCP servers to connect these systems while retaining production safeguards.
- Build shared infrastructure and organizational fluency. Model gateways, tool directories, managed deployment, authentication, and security controls help teams experiment without repeatedly rebuilding the same integrations. AI-enabled onboarding, cross-team technical communities, internal collaboration, and leadership workshops address uneven adoption across engineering organizations.
Zhang also situates Bloomberg’s infrastructure work within open-source ecosystems including KServe and Envoy AI Gateway, without claiming personal ownership of either project. His defining argument is that generative AI changes the cost function of software engineering: previously uneconomical maintenance becomes feasible, while review, integration, and assurance become more consequential.
