Preparation for Becoming requires a shift from "how do I optimize this" to "how do I verify this." Your interviewers are looking for a deep, first-principles understanding of why models behave the way they do.

• Role-Related Knowledge – You must demonstrate a rigorous grasp of dynamical systems, control theory, and sequence modeling. Be prepared to discuss the mathematical underpinnings of your past work and how those principles apply to long-horizon prediction.
• Problem-Solving Ability – Becoming values candidates who can design experiments in the absence of clear benchmarks. Focus on your methodology: how you formulate hypotheses, design stress tests, and translate failure modes into actionable architecture changes.
• Communication of Ambiguity – You will be evaluated on your ability to articulate complex trade-offs. If a model has a fundamental limitation, your interviewer expects you to be the first to point it out and explain the implications for the broader system.

The interview process at Becoming is designed to mirror the actual work: it is high-trust, technical, and focused on your ability to operate in an ambiguous environment. You can expect a series of deep-dive technical discussions with research and engineering leads, a potential take-home or whiteboard design task, and cultural alignment conversations.

The rigor is high, but the process is intended to be a collaborative exploration of your research philosophy. We look for candidates who are intellectually honest about the limits of their models and who demonstrate the maturity to handle open-ended, foundational problems.

This visual timeline tracks your progression from initial screening to final technical evaluation. Use this to pace your study of dynamical systems and model validation strategies, ensuring you have enough time to review your own past research for deep-dive discussions.

Model Validation & Stress Testing

This is the heart of the role. We need to know if you can "break" a model to understand its true utility. Strong performance involves a structured approach to identifying edge cases.

• Experimental design – How you set up baselines when standard metrics fail.
• Failure mode analysis – Your ability to pinpoint exactly where and why a model loses predictive power.
• Stability and Generalization – Understanding the mathematical constraints of your models in long-horizon regimes.

Be ready to go over:

• Designing benchmarks for non-standard, time-evolving data.
• The distinction between overfitting to noise and capturing true underlying dynamics.
• How you iterate on a model based on failure analysis.

Technical Depth in Dynamics

We are looking for a strong theoretical foundation in dynamical systems. You should be comfortable moving between theoretical math and applied machine learning.

• State-space models & Neural ODEs – Knowledge of architectures designed for continuous-time or long-horizon tasks.
• Control theory – Understanding how to influence or predict trajectories in complex systems.
• Hybrid approaches – Integrating physical or biological priors with learned models.

As an AI Research Scientist, you will own the validation lifecycle of our core modeling primitive. Your day-to-day will involve designing experiments that push the boundaries of current AI capabilities, moving beyond standard benchmarks to test stability, generalization, and predictive fidelity.

You will work closely with engineering teams to translate your research findings into actionable design changes. This means you aren't just writing papers; you are shaping the architecture of the platform. You will be expected to explore applicability across multiple domains, ensuring that our Developmental Intelligence is as robust in physical systems as it is in biology.

We are looking for candidates who combine academic rigor with the high-agency mindset of an early-stage company.

• Must-have skills:
• PhD or equivalent experience in applied math, physics, CS, or machine learning.
• At least 1 year of industry experience with real-world modeling systems.
• Demonstrated experience in dynamical systems, control, or sequence modeling.
• Ability to design validation strategies for partial, delayed, or noisy ground truth.
• Nice-to-have skills:
• Experience with world models, diffusion over time, or hybrid architectures.
• Exposure to biological or physical systems.
• A track record of identifying why models fail, not just improving performance metrics.

Q: How difficult are the technical interviews? A: They are challenging and research-oriented. Expect deep-dive questions that require you to think from first principles rather than relying on common industry heuristics.

Q: What is the company culture like? A: Becoming is a high-trust, high-ownership environment. We value direct communication, intellectual honesty, and the ability to work on foundational problems with ambiguous outcomes.

Q: What is the typical timeline for an offer? A: While it varies, we aim to be efficient. From initial screen to offer, the process usually spans a few weeks, depending on your availability and the number of stakeholders involved.

The provided salary data reflects the broad range of compensation for this role, which includes a mix of base salary and equity. Candidates should view this range as an indicator of the high-impact nature of the position and our commitment to competitive, long-term alignment.

• Focus on the "Why": In every technical interview, explain the reasoning behind your architectural choices. We care more about your process than the specific tools you used.
• Own the Ambiguity: When given an open-ended problem, acknowledge the unknowns. A great candidate defines the scope of the problem before jumping into the solution.
• Be Direct: If you think an approach is flawed, say so. We value the integrity to speak up when something won't work.
• Prepare for Deep Dives: Be ready to explain the most complex project you have worked on in minute detail, including the specific mathematical challenges you faced.

The AI Research Scientist role at Becoming is an opportunity to build the foundation of Developmental Intelligence. You will be directly responsible for the rigor and validation of a new modeling primitive that aims to solve some of the most difficult problems in AI today.

Your preparation should focus on your ability to think deeply about dynamical systems and your willingness to rigorously stress-test your own work. By focusing on first principles, experimental design, and clear communication, you will be well-positioned to succeed. We encourage you to review your past research projects through the lens of failure analysis and system stability. Good luck—you are preparing to contribute to a mission that could redefine how we model the world.