Senior Technical Product Manager - AI Agents, Evals & Reliability

About the role

A1 builds a proactive smart assistant for everyday users, aiming for high reliability in long-running workflows, persistent context and real-world task completion, even with non-deterministic model behavior. This is a deeply technical, hands-on role in which you work directly with engineers on system design, evaluation and trade-offs, turning AI potential into reliable behavior in a real-world application.

Key responsibilities

• Research and define end-to-end AI system requirements from capability to behavior to user impact.
• Translate model capabilities, data constraints and evaluation results into clear product and system decisions.
• Make hard trade-offs across quality, latency, cost, reliability and UX.
• Work closely with ML, backend and mobile engineers on system design, evaluation and iteration.
• Define and evolve evaluation frameworks across offline metrics, online experiments and human feedback.
• Drive execution with clear specs, strong judgment and disciplined prioritization, ensuring systems ship quickly, safely and reliably.
• Own product quality end-to-end: correctness, predictability and user trust.

Required profile

• Significant experience owning complex, technical products end-to-end.
• Proven ability to work closely with senior engineers and ML teams.
• Strong judgment and decision-making in ambiguous, fast-moving environments.
• Nice to have: experience shipping AI-heavy consumer products, background as an engineer or highly technical product manager, experience defining evaluation metrics for ML systems, and prior experience in zero-to-one product environments.

Required skills

• Strong grounding in computer science fundamentals, including algorithms, data structures and system design.
• Solid understanding of ML fundamentals and how modern AI systems behave in production.
• Comfort reading, reviewing and discussing technical design documents.
• Hands-on exposure to AI-powered products, including LLM-based systems, model evaluation, prompt or pipeline iteration and feedback loops.