Applied Research Engineer, Post-Training & Environments

Embedding VC; New York (NY); On-site; USD 140,000 - 210,000; Full time

Job summary

Chakra Labs in New York City is seeking a focused ML engineer to advance post-training loops, environment design, and high-fidelity data collection. You will help shape model improvement workflows across supervised fine-tuning, preference data, and reinforcement learning approaches, contributing to meaningful signals that actually improve models.

The role blends research curiosity with production pragmatism, with responsibilities spanning training infrastructure, evaluation harnesses, and

Qualifications

• Masters or PhD-level knowledge of machine learning fundamentals
• Hands-on post-training experience with LLM fine-tuning, preference data, or reinforcement learning for language models
• Environment-building instinct for agent environments, eval harnesses, and production-ready tools
• Python and PyTorch fluency; ability to work across ML infra like FastAPI

Responsibilities

• Design and run post-training loops across supervised fine-tuning, preference optimization, and reinforcement learning for language models
• Build environments and tasks, validators, reward signals, and evaluation harnesses that test meaningful capabilities
• Create, inspect, and improve data trajectories for training frontier models
• Develop reward functions, rubrics, validators, and analysis tools to turn model behavior into useful signals
• Contribute to distributed GPU training, PyTorch/FSDP, and ML infrastructure for training and data generation
• Translate ambiguous research goals into concrete environments, datasets, experiments and deliverables

Job description

Chakra Labs in New York City is seeking a focused ML engineer to advance post-training loops, environment design, and high-fidelity data collection. You will help shape model improvement workflows across supervised fine-tuning, preference data, and reinforcement learning approaches, contributing to meaningful signals that actually improve models.

The role blends research curiosity with production pragmatism, with responsibilities spanning training infrastructure, evaluation harnesses, and