Research Scientist

Seldon Technologies, Inc; San Francisco (CA); On-site; USD 150,000 - 210,000; Full time

Benefits offered by this job

Visa sponsorship

Relocation support

Professional development budget

Team off-sites

Conference attendance

Transportation cost support

Gym membership

Food stipend

Job summary

Seldon Technologies, Inc. is seeking a research-focused professional to advance multimodal data generation, post-training experiments, and data quality through careful design and evaluation.

You will collaborate with researchers, research engineers, and the evaluations team to propose projects, run experiments, and interpret results. You will bring hands-on ML model training or fine-tuning experience, strong Python skills, and familiarity with PyTorch or JAX.

Qualifications

• Experience training or fine-tuning ML models with a focus on optimization and evaluation.
• Ability to design rigorous experiments and interpret uncertain results.
• Strong Python skills with hands-on experimentation in PyTorch, JAX, or similar frameworks.
• Proven track record of research or technical projects from academia, industry or independent work.

Responsibilities

• Develop and test approaches to multimodal data generation, curation, and post-training.
• Design fine-tuning and reinforcement learning experiments, selecting baselines and varying factors to understand effects.
• Inspect model outputs and agent actions to identify failures and plan next investigations.
• Evaluate performance on unfamiliar tasks and document methods, findings, and limitations for reuse.

Skills

Python

PyTorch

JAX

Experiment design

Data analysis

Collaboration

Research projects

Education

PhD preferred

Tools

Hugging Face

Transformers

vLLM

TRL

Job description

You’ll conduct research on multimodal models, post-training, and data quality, with the freedom to explore other directions relevant to Seldon. We expect you to propose your own projects and contribute to shared research, working with researchers, research engineers, and the evaluations team to develop ideas, design and run experiments, and interpret the results.

What you’ll work on

• Develop and test approaches to multimodal data generation, curation, and post-training.
• Design fine-tuning and reinforcement learning experiments, choosing baselines and varying individual factors to understand their effects.
• Inspect model outputs and agent actions to identify failures and decide what to investigate next.
• Evaluate performance on unfamiliar tasks and document methods, findings, and limitations so others can build on the work.

What you’ll bring

• Experience training or fine-tuning machine learning models, with an understanding of optimization and model evaluation.
• Strong Python skills and experience implementing experiments in PyTorch, JAX, or a similar framework.
• Careful experiment design and data analysis, including choosing meaningful comparisons and interpreting uncertain results.
• Research or substantial technical projects from academic, industry, or independent work.
• You take initiative and follow through on your work. You’re principled in your decisions, open about uncertainty, and collaborative in how you solve problems.

Nice to have

• Experience with multimodal models, video understanding, synthetic data, or reinforcement learning for agents.
• Familiarity with Hugging Face Transformers and Datasets, training libraries such as TRL, or inference tools such as vLLM.
• A PhD in a relevant field and research publications are preferred, but neither is required.
• Comprehensive health, dental, and vision coverage.
• Visa sponsorship and relocation support.
• A professional development budget, the best AI tools, GPUs, a Mac, and the subscriptions you need.
• Team off-sites and conference attendance.
• Support with transportation costs.
• Gym membership and sports club subscriptions to support your wellbeing.
• A food stipend with nearly full meal coverage.