Research Co-Founder at Voltaire

As a Research Co-Founder at Voltaire, you will own the training agenda for Volt, a decision model that operates companies. You will define training objectives, design post-training on sparse signals, and build continual-learning loops. This is a founding seat for a world-class researcher to define an entirely new category of autonomous AI.

• Design and implement post-training for Volt using offline reinforcement learning on logged business decisions and their sparse, long-horizon measured outcomes.
• Build the continual-learning infrastructure that allows the model to improve from real-world results without leaking sensitive data between different business tenants.
• Define the core training objectives and credit assignment strategies that allow a model to navigate confounded signals to optimize for measurable business growth.

• Proven frontier lab experience at DeepMind, OpenAI, or Anthropic, or a PhD from a top-tier group like Stanford or MIT with first-author papers at NeurIPS, ICML, or ICLR.
• Deep technical expertise in reinforcement learning, offline RL, or self-improvement of models, with identifiable public contributions and active training code experience within the last year.
• Exceptional first-principles reasoning with the ability to derive new training methods and state precisely what a signal supports and what it does not in a whiteboard environment.

• True founding ownership: You own the research agenda and policy architecture outright, working directly with a founder who scaled his first company to $15 million in year one.
• High-stakes research: Voltaire is raising $10M+ from Tier 1 VCs to move beyond reactive LLMs toward autonomous, goal-oriented business policies capable of long-horizon reasoning.
• Unrivaled impact: You aren’t calling model APIs; you are deriving the methods to train models that estimate business states and execute actions through Slack and systems of record.