SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
DoubleVerify is seeking a Senior Manager of Engineering to lead the PinAI team, a strategic initiative building conversational AI and agentic systems for their Pinnacle analytics platform. This role combines technical leadership with people management, overseeing the development of AI-powered interfaces that help Fortune 500 clients interact with advertising data through natural language.
You will define the technical vision and roadmap for PinAI, architecting agentic AI systems using Model Context Protocol (MCP), LangChain, LangGraph, and advanced LLM capabilities. Key responsibilities include building and scaling a world-class team of AI/ML engineers and technical leads, establishing technical standards for AI/ML engineering (evaluation frameworks, prompt engineering, model optimization), and ensuring production-grade quality through comprehensive testing and observability.
On the technical side, you'll oversee development of conversational AI interfaces with context awareness and conversation history, implement robust evaluation frameworks to measure AI agent performance, architect modular AI components using MCP, and integrate with Pinnacle's existing analytics infrastructure (Looker, data APIs, authentication). You'll also drive MLOps best practices, optimize LLM API costs, and establish incident response protocols and SLAs.
Cross-functional collaboration is essential: you'll partner with Product Management on AI-powered features, work with Data Science on model selection and fine-tuning, coordinate with Platform Engineering on LLM deployment infrastructure, engage with Security and Compliance on responsible AI practices, and collaborate with UX/UI teams on seamless conversational interfaces.
You'll also own operational excellence, defining and tracking KPIs for AI system performance (accuracy, latency, user satisfaction, engagement), implementing MLOps best practices, and driving continuous improvement through retrospectives and user feedback loops.
