Senior ML Researcher: Large Models, Real-World RL, Equity

raydar; San Francisco (CA); On-site; USD 200,000 - 300,000; Full time

Benefits offered by this job

Health, medical and dental coverage

Job summary

Raydar in San Francisco, CA seeks a Machine Learning Researcher to lead research from hypothesis to working models, scaling large neural networks and aligning offline experiments with real-world results. The role emphasizes hands-on model training, data pipelines and evaluation design.

You will train large models, design reinforcement learning methods, build synthetic data pipelines, run distributed training, and mentor colleagues while publishing or releasing open-source work.

Qualifications

• 5–10 years in research or engineering roles; doctoral study counts toward experience.
• Proven ability to train and refine large neural networks with clear justification of results.
• Fluent coding in a Python-based deep learning stack for research.
• Experience with distributed training frameworks and scalable pipelines.

Responsibilities

• Modify and train large models, including base-training passes and instruction tuning.
• Build reinforcement learning methods that learn from real-world outcomes with noisy feedback.
• Design synthetic-data and knowledge-distillation pipelines with provenance tracking.
• Create benchmarks and evaluations to test generalization and connect offline results to practice.
• Run and troubleshoot large distributed training jobs, improving throughput and cost efficiency.
• Mentor colleagues, set research agenda, and share work via papers or open-source releases.

Skills

Distributed training

Deep learning research

Python for research

Model evaluation & ablation

Research leadership

Education

PhD in CS/ML

Tools

FSDP

DeepSpeed

Megatron

PyTorch

Job description

Raydar in San Francisco, CA seeks a Machine Learning Researcher to lead research from hypothesis to working models, scaling large neural networks and aligning offline experiments with real-world results. The role emphasizes hands-on model training, data pipelines and evaluation design.

You will train large models, design reinforcement learning methods, build synthetic data pipelines, run distributed training, and mentor colleagues while publishing or releasing open-source work.