Adversarial ML Scientist: Robustness & Efficient Models

Mercor; San Francisco (CA); On-site; USD 180,000 - 260,000; Full time

Benefits offered by this job

Flexible, project-based work

Competitive compensation

Job summary

Obsidian is seeking experienced ML researchers to tackle end‑to‑end deep learning challenges across vision and language. You will train image classifiers, generative models, and open‑weight LLMs, while pushing robustness, efficiency, and deployment‑ready performance on high‑scale data.

Applicants should have 3+ years in ML research, expertise in PyTorch/JAX/TensorFlow, and a strong publication or open‑source track record from a top university or FAANG‑level environment.

Qualifications

• 3+ years of machine learning research experience (PhD counts)
• Strong experience with PyTorch, JAX, TensorFlow or similar ML frameworks
• Degree from a top‑100 university or equivalent research impact

Responsibilities

• Train image classifiers and generative image models end‑to‑end
• Fine‑tune open‑weight language models and optimize robustness
• Diagnose and resolve training issues and latency constraints
• Compress models to meet size and speed budgets
• Evaluate and iterate using sample‑quality metrics like FID

Skills

Adversarial robustness

Efficient computer vision

Generative image modeling

LLM fine-tuning

Multilingual pre-training

Data augmentation

Education

PhD or equivalent research track

Top‑100 university degree

Tools

PyTorch

JAX

TensorFlow

Job description

Obsidian is seeking experienced ML researchers to tackle end‑to‑end deep learning challenges across vision and language. You will train image classifiers, generative models, and open‑weight LLMs, while pushing robustness, efficiency, and deployment‑ready performance on high‑scale data.

Applicants should have 3+ years in ML research, expertise in PyTorch/JAX/TensorFlow, and a strong publication or open‑source track record from a top university or FAANG‑level environment.