VP of Research (Machine Learning)

Lead the intelligence and research direction behind a proactive AI assistant designed to reduce time spent on routine work through reliable, multi-step, tool-using AI. The product supports high-frequency use and focuses on dependable long-running workflows, persistent context, and task completion.

Responsibilities

• Define the research roadmap for memory, context handling, reasoning, planning, and orchestration in the assistant’s core intelligence.
• Decide when to build custom model architectures versus adapting open-source or commercial frontier models.
• Build evaluation systems that measure real-world reliability and safety beyond benchmark scores.
• Make alignment, safety, and guardrails core product decisions.
• Explore retrieval-augmented training, mixture-of-experts, distillation, multi-agent systems, and multimodal input.
• Partner with product and engineering to shape the assistant’s capabilities early.

Requirements

• Fluency in Python and PyTorch or JAX, with comfort running GPU training and inference at scale.
• Experience shipping or evolving ML systems in production rather than only research prototypes.
• Strong judgment about model failure modes, behavior, and long-term reliability.
• Hands-on, real-world building mindset.
• Ability to make high-stakes decisions with incomplete information.
• Strong focus on evaluation, correctness, and system behavior over time.
• Founder-level ownership and willingness to lead directly.

Nice to Have

• Experience building a startup research function from zero to one.
• Practical experience with retrieval-augmented training, MoE, distillation, multi-agent systems, or multimodal models.
• Experience owning safety and guardrail strategy for a live product.

What They Offer

• Cash and equity compensation.
• Remote-first work with flexible hours on a distributed global team.
• Generous paid time off.
• Company laptop.
• A fast hiring process with three, and occasionally four, interviews followed by prompt decisions.

About the Role

The work centers on a proactive assistant for everyday productivity workflows such as email, notes, and tasks. Its emphasis is reliability in long-running workflows, persistent context, and getting tasks completed despite the non-deterministic nature of underlying models.