Senior AI Researcher

About the role

The Biological Computing Co. (TBC) develops AI models inspired by biological neural networks to improve computational efficiency. We are seeking a researcher to build video generation systems that allow robots to plan and act by imagining future scenarios.

Key facts

What you'll do

• Architect video generation models that feature stable rollouts and controllable latent representations.
• Enhance the fidelity of long-horizon autoregressive predictions beyond simple one-step accuracy.
• Incorporate physical structure and object-centric data into control systems.
• Analyze the balance between inference cost, latency, and model performance for robotic applications.
• Manage research projects from the initial hypothesis through to system-level implementation and testing.
• Detect and mitigate scaling or training risks before they impact development.
• Collaborate with founders, biologists, and engineers to turn neural principles into functional software.
• Provide technical mentorship and guidance to other team members.

Requirements

• Background in robotics, machine learning, or computer vision.
• Direct experience designing and training generative models, specifically diffusion, sequence models, or autoregressive video.
• Experience with physics-informed learning, model-based reinforcement learning, or system identification.
• Ability to manage the failure modes associated with long-horizon, autoregressive rollouts.
• Experience navigating the full research lifecycle, including architecture design, training, and evaluation.
• Demonstrated ability to make high-level architectural decisions based on first-principles thinking.
• Ability to communicate technical trade-offs regarding compute, quality, and control utility.

Nice to have

• PhD or MS in Robotics, Machine Learning, or Computer Science.
• Research or industry history in embodied AI, learned simulation, or generative video.
• Experience training policies using imagined trajectories or learned simulators.
• Familiarity with action-conditioned video prediction or controllable generative models.
• Experience deploying learned models on physical robotic hardware.
• Knowledge of cross-embodiment learning, latent-action models, or learning from human video data.
• Experience with structured dynamics models, physical priors, or object-centric representations.
• Experience with sim-to-real transfer, digital twins, or online adaptation.
• Experience scaling systems across distributed training environments or large datasets.
• Publication record at major machine learning, robotics, or computer vision conferences.

Skills & tools

• Generative models (diffusion, autoregressive)
• Model-based reinforcement learning
• Physics-informed learning
• Robotic control systems
• Distributed training
• System identification

Practical notes

Our interdisciplinary team includes researchers and engineers with backgrounds from Apple, Johns Hopkins, Meta, MIT, and Stanford.