Research Scientist, Dynamical Systems & AI

Becoming•San Francisco, CA

13h

About The Position

Becoming is building Developmental Intelligence: AI for predicting how organisms change over time. Most existing models work in short-horizon or static regimes. They fail when systems become long-horizon, nonlinear, and context-dependent — exactly where development, biology, and many real-world systems live. We are building a new modeling primitive for reasoning about complex dynamics over time, starting with developmental biology and extending to other domains where prediction fundamentally breaks. We are hiring a Research Scientist to help verify, validate, and shape a new modeling primitive designed for complex, time-evolving systems. This role is not about incremental model improvements or benchmark chasing. Your responsibility is to rigorously evaluate whether this primitive works, where it works, where it fails, and why. You will help define its scope, limits, and evolution through careful analysis, experimentation, and comparison to existing approaches. Your work will directly influence how the platform develops and how broadly it can be applied.

Requirements

PhD or equivalent experience in applied mathematics, physics, computer science, machine learning, or a related field
At least 1 year of industry or applied research experience working on real modeling systems
Strong grounding in dynamical systems, time-series modeling, control, or sequence modeling
Experience evaluating models where ground truth is partial, delayed, or noisy
Ability to design validation strategies when standard benchmarks are insufficient
Comfort working with first-of-its-kind architectures and open-ended questions

Nice To Haves

Experience with state-space models, world models, neural ODEs, diffusion over time, or hybrid approaches
Prior work on systems where prediction degrades over long horizons
Exposure to biological, physical, or other real-world dynamical systems
Track record of identifying why models fail — not just improving metrics

Responsibilities

Validation of a new AI modeling primitive for long-horizon dynamical systems
Design of experiments and benchmarks that test stability, generalization, and predictive fidelity over time
Comparative evaluation against existing modeling approaches
Identification of failure modes, assumptions, and edge cases
Clear articulation of why the model succeeds or fails in different regimes
Translation of findings into guidance for future architecture and system design
Exploration of applicability across multiple domains with complex dynamics, not just biology