Graduate Research Scientist: Seed Model Alignment & Safety

ByteDance; San Jose (CA); Vor Ort; USD 254.000 - 480.000; Vollzeit

Benefits dieser Stelle

Medical insurance

401(k) with company match

Paid parental leave

Zusammenfassung

ByteDance is seeking a Research Scientist Graduate (Seed Model-Alignment & Preparedness) for a 2027 start in San Jose. You will contribute to safety-aligned model development, RL-based safety training, and risk assessment across high-stakes domains.

Strong ML foundations and hands-on experience with LLMs, RL, and Python/PyTorch are essential. Responsibilities include automated red teaming, monitoring, and scalable deployment, collaborating with cross-disciplinary teams, and publishing results to

Qualifikationen

• PhD or near completion in CS/EE/EC Eng/Physics/Math or related field.
• Strong foundations in ML, DL, statistics with experience in LLMs/RL or related.
• Proficiency in Python and PyTorch with ability to implement algorithms and build experiments.
• Hands-on experience in agent RL, model training, dataset development, or ML infrastructure.
• Ability to translate open-ended model questions into testable hypotheses, run experiments, and report results.

Aufgaben

• Agent RL & safety training: design environments, rewards, and data to minimize deception and ensure safe autonomy.
• Frontier risk assessment with domain experts to evaluate dangerous capabilities and safeguards.
• Automated red teaming: build attack generation and adversarial training to harden models.
• Monitoring & oversight: develop oversight models, permission controls, and evaluate effectiveness.
• Failure analysis & deployment: transform real-world failures into reproducible experiments and scale improvements.

Kenntnisse

ML Foundations

RL

LLMs

PyTorch

Python

Experiment pipelines

AI safety

Safety in AI

Ausbildung

PhD in CS/EE/EC Eng/Physics/Math

Tools

Python

PyTorch

Jobbeschreibung

ByteDance is seeking a Research Scientist Graduate (Seed Model-Alignment & Preparedness) for a 2027 start in San Jose. You will contribute to safety-aligned model development, RL-based safety training, and risk assessment across high-stakes domains.

Strong ML foundations and hands-on experience with LLMs, RL, and Python/PyTorch are essential. Responsibilities include automated red teaming, monitoring, and scalable deployment, collaborating with cross-disciplinary teams, and publishing results to