The Key Responsibilities Of This Role Are

• Contribute to RL frameworks that drive the design-make-test-analyze (DMTA) cycles that power our EMMI platform, which coordinates a closed-loop between a highly automated lab and our reward models.
• Develop synthetic data engines and the inference infrastructure needed to simulate environments for large-scale training.
• Maintain rigorous evaluations to continually monitor the performance of learned policies, using large proprietary datasets collected from internal programs.

Experience and Qualifications: Part of Terray’s success is nurtured by a hands-on work environment where everyone is accountable, vested in a vision of excellence, and actively taking part in the success of the business. Terray supports a positive work environment where employees can feel engaged, recognized and empowered to be creative.

Required Qualifications

• Strong experience in machine learning engineering, with interest in techniques for sequential decision-making: bayesian and black-box optimization, reinforcement learning.
• Ability to quickly switch between robust engineering and exploration of conceptual insights, e.g., implementation details of training on asynchronous rollouts while understanding why policy divergence leads to instabilities.
• Experience with the challenges of complex real-world systems and scientific environments, such as expensive queries and experimental noise.
• Appreciation for elegant ideas and what works in practice.

Preferred Qualifications

• Experience with synthetic data for chemistry, frameworks for autonomous discovery, test-time training.

Only applicants with github, proof of relevant work, or a one-page writeup of experience applying autonomous discovery to a scientific problem that is verifiable will be considered. Compensation Details: $147,000 - 227,850 (annually) depending on experience; participation in the Company's option plan; 3% retirement safe harbor contribution; fully-paid medical, dental, vision, life and disability benefits and much more.