🛠 Tech Stack

💼 About This Role

You'll study how frontier models acquire durable behavioral tendencies across the training stack. You'll define target behaviors, design data and training interventions, and build evaluation loops to determine whether learned behaviors are broad and robust. This role is close to the core training loop for a leading AI research company.

🎯 What You'll Do

• Develop synthetic data methods for training behavioral tendencies
• Study how pre-training, mid-training, and post-training shape behavior
• Build evaluation loops connecting behavior to training data
• Design reusable data generation and filtering pipelines
• Create experiments distinguishing durable behavior from artifacts

📋 Requirements

• Record of technically excellent work in large-scale ML
• Experience with pre-training, post-training, synthetic data, or model evaluation
• Ability to design experiments with subtle or noisy signal
• Strong judgment about research questions worth pursuing

✨ Nice to Have

• Alignment research background
• Experience with training infrastructure
• Cross-functional collaboration skills

🎁 Benefits & Perks

• 💰 Competitive salary and equity
• 🏖️ Flexible PTO
• 🏥 Health insurance
• 📚 Learning and development budget

📨 Hiring Process

Estimated timeline: 3-5 weeks · AI estimate

1. 1Recruiter Call· 30 min
2. 2Technical Interview· 60 min
3. 3Research Presentation· 60 min
4. 4Hiring Committee· 30 min