Preeti Somal is executive vice president of Build, Operate, and Secure at Temporal, leading engineering, product, and design. Her career spans Oracle, VMware, Yahoo, HashiCorp, and Temporal, with a sustained focus on the infrastructure and engineering practices that keep complex applications reliable in production.
From cloud infrastructure to dependable AI
Somal studied electrical and electronics engineering at Punjab Engineering College and earned a master’s degree in computer science from the University of Georgia. She began her career at Oracle before moving into engineering leadership at VMware, where she worked on application and systems management.
As Yahoo’s vice president of cloud services from 2013 to 2017, she led teams operating internal infrastructure for its consumer and advertising businesses. She joined HashiCorp as vice president of engineering in 2018 and became executive vice president of engineering in 2021, overseeing engineering, security, and IT. As HashiCorp expanded, she helped develop its remote-first engineering culture, organizing teams around product workstreams and embedding shared engineering services to remove operational bottlenecks.
AI agents are distributed systems. Agents coordinate model calls, tools, conversation history, parallel tasks, and human approvals; each dependency can fail. Somal applies established distributed-systems principles to making these applications reliable in production.
Durable execution separates business logic from recovery. Developers write workflows using language-native SDKs, including Python, while Temporal manages retries, persistent state, and failure recovery. Application workers remain in the developer’s environment; Temporal Cloud supports execution and reliability without replacing existing CI/CD practices.
Execution history makes agents inspectable. Persistent, exportable workflow records help engineers reconstruct tool calls, debug failures, test behavior, and address compliance requirements. Signals and queries let long-running agents accept new input and expose their current state.
Blameless incident response strengthens reliability. Somal advocates accessible runbooks, open incident reviews, leadership participation during outages, and anonymous feedback from on-call engineers. Her emphasis on psychological safety for on-call engineers positions organizational learning as essential production infrastructure.
