New: upload your resume and get matched to AI jobs. Free to start →

Post-Training Research Scientist - Forecasting and Reasoning

New York • Full-time

Posted 9d ago

Remote Work Policy

On-site

Employment Type

Full-time

Categories

LLM Engineer

About the job

We develop and utilize Large Language Models (LLMs) for forecasting and decision-making within financial markets. Our research integrates structured time series data, unstructured information, and computational tools to understand how models acquire evidence, make calibrated predictions, and adapt their beliefs with new information. We leverage proprietary datasets, domain expertise, and real investment applications to assess these capabilities, working with rich, challenging, noisy, and non-stationary data. Our focus is on advancements in training environments, reward design, policy optimization, post-training techniques, and evaluation methodologies.

Responsibilities

  • Develop post-training methods for forecasting and reasoning, including supervised fine-tuning and reinforcement learning, to enhance accuracy and calibration.
  • Build point-in-time environments that combine historical observations, documents, prior predictions, and tools, and define tasks requiring information gathering and belief updating.
  • Develop reward modeling and credit assignment methods using verifiable outcomes, expert supervision, preference data, and delayed feedback.
  • Build efficient and reliable training systems, collaborating with engineers on synchronous and asynchronous learning pipelines for multi-node GPU clusters.
  • Establish rigorous evaluations to measure forecasting accuracy, uncertainty calibration, reasoning quality, robustness, and downstream utility.
  • Productionize research by collaborating with quantitative researchers to identify valuable model capabilities and deploy successful methods into production systems.

Requirements

  • Bachelor’s degree or higher in a quantitative or technical discipline, or equivalent practical experience in advanced learning systems.
  • Demonstrated expertise in LLM post-training, reinforcement learning, environment design, or preference learning; experience in non-fully verifiable domains is preferred.
  • Hands-on experience with distributed GPU training, including performance optimization, numerical stability, checkpointing, and reproducibility.
  • Strong implementation skills in Python and PyTorch or JAX for research and production model training.
  • Experience using distributed training and inference frameworks such as FSDP, DeepSpeed, SGLang, Megatron, verl, or Slime.
  • Ability to design rigorous evaluations for forecasting accuracy, calibration, reasoning quality, robustness, temporal holdouts, and leakage detection.
  • Experience with probabilistic forecasting, time-series modeling, asynchronous reinforcement learning (RL), tool-using agents, or production post-training systems from academic research or AI labs.
  • Commitment to high ethical standards.

Benefits

  • Fully-paid health care benefits
  • Generous parental and family leave policies
  • Volunteer opportunities
  • Support for employee-led affinity groups
  • Mental and physical wellness programs
  • Tuition assistance
  • 401(k) savings program with an employer match