Post-Training Research Scientist
A greenfield research role building RLHF, DPO and reward-modelling capability for large language models applied to financial and quantitative tasks. It sits in a quantitative investment firm's AI research effort.
What you’d work on
Our reading of the posting: what a new hire in this role would most likely spend their first months on.
- Build preference-data and reward-modelling infrastructure on financial datasets
- Run RLHF and DPO experiments at large compute scale
- Define evaluation frameworks with quant researchers
- Support production systems that depend on post-trained models
What they ask for
Required
- Shipped post-training systems such as RLHF or DPO
- Distributed GPU training
- Managing large compute budgets and ablations
- Reward modelling or alignment expertise
- PyTorch or JAX
- DeepSpeed or FSDP
Preferred
- Master's degree
- Experience at a frontier AI lab
- Education
- Bachelor's in a STEM field or equivalent work experience; Master's a plus.
- Experience
- At least 1 year; 1–5 years ideal, at a frontier AI lab or equivalent
- Languages and tools
How to prepare
- Fine-tune a small language model with DPO and compare it to supervised fine-tuning
- Learn distributed training with FSDP or DeepSpeed and its failure modes
- Study reward modelling when feedback is delayed and noisy
- Design an evaluation set for quantitative reasoning tasks
What to study
- Financial Machine Learning
- Text as Data and NLP for Finance
- Python for Quantitative Finance
- Probability, Statistics and Financial Data
The NLP and ML Quant Researcher roadmap sets out the stages for this kind of role, with free resources for each.
Open the roadmapApply
Read the full posting on Two Sigma’s site before applying: requirements, location and pay can change after we check them.
View the posting on Two Sigma’s site ↗
SelfMadeQuant wrote this summary from the employer’s public posting, last checked 6 November 2026. SelfMadeQuant isn’t affiliated with Two Sigma and doesn’t handle applications.