LLM & Computer Vision Researcher at Mercor

LLM & Computer Vision Research Scientist | $100–120/hr | Worldwide Remote

Mercor is seeking experienced machine learning researchers with hands-on expertise training and improving deep learning models end-to-end across vision and language. You'll tackle well-scoped, empirical, open-ended ML research problems with real-world impact.

Responsibilities

• Train image classifiers and generative image models from scratch; fine-tune open-weight language models.
• Maximize performance under limited data, compute, and model-size budgets.
• Harden models against adversarial inputs and adversarial conversations.
• Compress models to meet strict size and latency constraints without sacrificing accuracy.
• Diagnose and resolve training instabilities and performance issues.

Key Areas of Expertise

Adversarial Robustness:

• Adversarial training of image classifiers (PGD, TRADES).
• Evaluating robust accuracy under standard threat models (L∞, AutoAttack) and avoiding gradient-masking pitfalls.
• Managing robustness–accuracy trade-offs and robust overfitting.

Efficient Computer Vision:

• End-to-end training for fine-grained recognition tasks.
• Model compression: quantization, pruning, and knowledge distillation.
• Deployment under hard size or latency budgets (edge, on-device, embedded).

Generative Image Modeling:

• Training diffusion models, GANs, VAEs, or flow-based models from scratch.
• Iterating on sample-quality metrics such as FID.
• Training-efficiency techniques for fast, compact generators.

LLM Post-Training & Behavioral Robustness:

• Supervised fine-tuning and preference optimization (DPO, RLHF, RLAIF).
• Shaping multi-turn conversational behavior: resistance to sycophancy, calibrated confidence.
• Alignment-style fine-tuning that modifies specific behaviors while preserving general capability.

Multilingual Pre-training:

• Training multilingual or low-resource-language models from scratch.
• Tokenizer design across diverse scripts and languages.
• Balancing highly unequal per-language data distributions.

General Qualifications

• 3+ years of ML research experience (PhD research counts).
• Strong proficiency in PyTorch, JAX, TensorFlow, or similar frameworks.
• Degree from a top-100 university, experience at a FAANG or top AI company, or a strong publication/open-source track record.

Why Join

• Work on cutting-edge machine learning research problems.
• Collaborate with leading AI researchers on high-impact projects.
• Flexible, project-based engagement with competitive hourly compensation.