Seeking experienced Machine Learning and NLP professionals to design expert-level AI evaluation tasks, develop reference solutions, and evaluate frontier AI models for a leading AI laboratory.
LLM Research Scientist (Pre-training & Post-Training)
Job description
Mercor is seeking experienced Machine Learning Researchers with hands-on expertise in training and improving large language models (LLMs) end-to-end. In this remote contract role, you'll tackle empirical, open-ended research problems involving foundation model pre-training, post-training, data curation, and model alignment for leading frontier AI labs.
Key Responsibilities
Foundation Model Training
- Train transformer-based language models from scratch.
- Fine-tune open-weight LLMs.
- Optimize model performance under limited compute and data budgets.
- Diagnose optimization failures, convergence issues, and training instability.
Pre-training Data Engineering
- Build large-scale training corpora from raw web crawls and other unfiltered datasets.
- Perform filtering, deduplication, quality classification, and data mixture optimization.
- Measure and evaluate the impact of data interventions.
LLM Post-Training
- Build supervised fine-tuning (SFT) datasets using synthetic generation, weak supervision, rejection sampling, and related techniques.
- Develop preference optimization pipelines (DPO, RLHF, RLAIF).
- Train reward models and human-preference predictors.
- Improve model alignment, truthfulness, refusal behavior, and reasoning quality.
- Fine-tune models for verifiable domains such as mathematics, programming, games, and structured prediction.
Research & Evaluation
- Investigate scaling laws and training efficiency.
- Explore curriculum learning and data ordering strategies.
- Design robust evaluation benchmarks and contamination controls.
- Conduct statistically rigorous model comparisons.
- Contribute to reinforcement learning for LLMs, alignment, and AI safety research.
Ideal Qualifications
Required
- 3+ years of machine learning research experience (PhD research counts).
- Strong experience with one or more of:
- PyTorch
- JAX
- TensorFlow
- Significant expertise in one or more areas:
- Foundation model pre-training
- Large-scale data curation
- LLM post-training
- RLHF/DPO/RLAIF
- Alignment research
Preferred
- Degree from a top-100 university.
- Experience at a FAANG company or comparable AI research organization.
- Strong publication record or impactful open-source contributions.
- Experience with:
- Scaling laws
- Curriculum learning
- Benchmark design
- Reinforcement learning for language models
- AI alignment and safety
Why Join
- Work on cutting-edge foundation model research.
- Collaborate with leading AI researchers.
- Flexible remote contract work.
- Competitive hourly compensation.
Compensation
- $100–$120 per hour
Job Details
- Remote
- Independent Contractor
- Flexible Schedule
- Weekly payments via Stripe or Wise
About Mercor
Mercor partners with leading AI labs and enterprises to develop frontier AI systems by leveraging expert human knowledge. Contributors work on high-impact research projects that advance the next generation of artificial intelligence.
You will be redirected to the company's website to complete your application.