Lead Product Manager

About the role

Dialpad is seeking a product leader to own the lifecycle of our core AI models, including custom SLMs, ASR stacks, and real-time inference infrastructure. This role is embedded in the technical layer of our platform, requiring deep involvement in model behavior, evaluation, and training data rather than standard administrative product management responsibilities. The hire will translate ambiguous quality issues into actionable metrics and clear release decisions while maintaining a technical peer stance during error analysis and incident retrospectives. You will define the data strategy that governs usage rights, consent, sampling, and annotation quality to ensure responsible model development. Success in this position depends on making informed trade-offs between model quality, streaming latency, infrastructure costs, and training methodologies under conditions of uncertainty. The role demands direct communication through written documentation, prioritizing memos and technical specifications over traditional slide decks to maintain alignment. You will set the roadmap for model development, including deprecation schedules and labeling priorities, while producing written direction memos that capture context and rationale. Ultimately, this position owns the full model lifecycle from data acquisition and training to production monitoring and retirement, ensuring that system outcomes align with business and user needs.

Key facts

What you'll do

• Manage the full model lifecycle from data acquisition and training to production monitoring and retirement, ensuring continuity and coherence across stages.
• Define data strategy, covering usage rights, consent, sampling, and annotation quality, to establish a reliable foundation for model training and evaluation.
• Translate ambiguous quality issues into actionable metrics and clear release decisions, balancing competing priorities under uncertainty.
• Participate in error analysis, incident retrospectives, and evaluation reviews as a technical peer, contributing expertise rather than only oversight.
• Support internal product teams by providing model capabilities and defining latency and cost constraints to ensure feasible and efficient implementations.
• Make trade-offs between model quality, streaming latency, infrastructure costs, and training methods, optimizing for user outcomes and business viability.
• Produce written decision documents, direction memos, and technical specifications that capture context, constraints, and rationale for future reference.
• Set the roadmap for model development, including deprecation schedules and labeling priorities, to guide focused and responsible investment.
• Interface closely with evaluation frameworks to design experiments that surface model weaknesses and validate improvements over time.
• Own the criteria for model evaluation, establishing thresholds and benchmarks that reflect real-world usage and risk tolerance.
• Coordinate with data annotation operations to ensure quality processes align with model requirements and regulatory expectations.
• Monitor production inference infrastructure to identify performance regressions and collaborate on remediation strategies.
• Engage with pricing and packaging considerations to ensure model capabilities are delivered in a sustainable and scalable manner.
• Act as the central point for questions regarding custom SLMs, ASR stacks, and real-time inference infrastructure, maintaining a current mental model of system behavior.

Requirements

• Hands-on experience building, training, fine-tuning, or optimizing models in a production environment, demonstrating familiarity with real-world constraints and failure modes.
• 2+ years of product ownership experience where you held accountability for system outcomes, showing that your decisions materially affected results.
• Technical fluency in evaluation design, quantization, serving trade-offs, and inference cost drivers, enabling informed collaboration with engineering teams.
• Ability to make decisions and commit to outcomes despite the probabilistic nature of model behavior, accepting that metrics may shift post-deployment.
• Direct communication style with a preference for written documentation over slide decks, ensuring clarity, precision, and efficient alignment across roles.
• Comfort working with ambiguous requirements and incomplete information while still driving toward actionable decisions.
• Experience interpreting error distributions and failure modes to guide prioritization and resource allocation.
• Willingness to dive into technical details without needing to be the author of every line of code, focusing instead on coordination and outcomes.

Nice to have

• Specific background in speech technology, including ASR, TTS, telephony, or streaming latency, providing domain context for model behavior and user expectations.
• Experience managing training data pipelines, labeling operations, or data rights, ensuring that data practices are sustainable and compliant.
• History of managing inference infrastructure at scale, including GPU capacity planning or cost-per-call optimization, to align technical and financial goals.
• Practical experience building or operating production evaluation harnesses that surface regressions and support rapid experimentation.
• Experience with the pricing or packaging of AI-driven products, helping to align model capabilities with market realities and customer value.

Practical notes

The role is full-time based in San Ramon, United States. Candidates must be eligible to work in the United States without sponsorship for this position. No visa sponsorship is available. The engagement is full-time, requiring standard business hours and readiness for occasional after-hours response during critical incidents. Travel is not expected as part of this role, given the primary reliance on remote collaboration and digital coordination.