Two Sigma
Post-Training Research Scientist
$165K–$300K
NY New York · Full-time · Posted Oct 28, 2026
- Experience
- Minimum 1 year required; 1–10 years preferred (ideally 1–5 years).
- Education
- BS or equivalent work experience in Science, Technology, Engineering or Math.
About the job
Build and lead post-training methods and infrastructure to align large language models with multi-step financial and quantitative workflows.
What you'll do
- Lead post-training efforts for LLMs applied to financial time series and quantitative reasoning, including RLHF, DPO, and related methods.
- Build preference-data, reward-modeling, and policy-optimization infrastructure; manage large-scale compute and experiments.
- Set research and evaluation direction with quantitative researchers and unblock production systems dependent on post-training.
What they ask for
- Shipped post-training systems in production, such as RLHF, DPO, or RLAIF.
- Deep understanding of distributed training infrastructure, including multi-node GPU clusters, training stability, and checkpointing.
- Track record managing large-scale compute, including budgeting, experiment design, and ablations.
- Publications or demonstrated expertise in alignment, preference learning, or reward modeling; hands-on PyTorch/JAX and distributed-framework implementation skills.
Nice to have
- 1–10 years of experience (ideally 1–5 years) at a frontier AI lab or equivalent.
- MS degree (a plus).
Pay
$165,000 to $300,000 USD base salary
Tools
RLHF, DPO, RLAIF, PyTorch, JAX, DeepSpeed, FSDP
AI focus
Develop post-training capabilities for LLMs, including RLHF, DPO, reward modeling, preference learning, and evaluation for financial and quantitative workflows.
Model-written summary · the original posting is authoritative