Staff Scientist, Post-Training & Evaluation Systems

Chakra Labs PVT LTD; New York (NY); On-site; USD 150,000 - 210,000; Full time

Job summary

Chakra Labs PVT LTD is seeking a research-forward engineer to advance post-training workflows, environment design, and evaluation infrastructure. You will work on creating high-fidelity trajectories and robust reward signals to guide frontier AI models through scalable training pipelines.

Ideal candidates have strong ML fundamentals, hands-on post-training experience, and proficiency with Python, PyTorch, and modern ML infra.

Qualifications

• Masters/PhD-level knowledge of machine learning fundamentals.
• Strong understanding of linear algebra, optimization, and gradient methods.
• Experience with LLM fine-tuning, reward modeling, or RL for language models.
• Fluency in Python; comfortable with PyTorch, FastAPI, and ML infra.
• Ability to bridge research ambiguity and production constraints.

Responsibilities

• Post-training loops: design and run model improvement workflows across supervised fine-tuning, preference optimization, and RL approaches like GRPO.
• Environment and task design: build environments and evaluative harnesses that test meaningful capabilities.
• High-fidelity trajectories: create, inspect, and improve data that teaches models how to behave.
• Reward and evaluation systems: develop rubrics, validators, and analysis tools to turn model behavior into training signals.
• Training and research infrastructure: run experiments on distributed GPU clusters and build scalable infra.
• Customer research problems: translate frontier research goals into concrete environments and deliverables.

Skills

ML fundamentals

Python

PyTorch

Environment design

Education

Master's or PhD in ML or related field

Tools

PyTorch

FastAPI

Distributed GPUs

FSDP

Job description

Chakra Labs PVT LTD is seeking a research-forward engineer to advance post-training workflows, environment design, and evaluation infrastructure. You will work on creating high-fidelity trajectories and robust reward signals to guide frontier AI models through scalable training pipelines.

Ideal candidates have strong ML fundamentals, hands-on post-training experience, and proficiency with Python, PyTorch, and modern ML infra.