Staff Machine Learning Research Scientist

What You’ll Do

• Lead end-to-end ML research, from idea generation to production deployment and monitoring
• Design, implement, and evaluate novel methods for LLM alignment on proprietary clinical data
• Develop and rigorously evaluate approaches for hallucination detection, attribution, and model reliability
• Build and curate high-quality datasets, with a strong emphasis on evaluation design and benchmark integrity
• Critically assess academic literature to identify strong vs weak methods, and translate the best ideas into practice
• Establish best practices for experimental design, including statistically sound evaluation and reproducibility
• Collaborate cross-functionally with engineering to productionize models (MLOps, infra, deployment)
• Develop methods for long-context and multimodal modeling (structured + unstructured clinical data)
• Mentor other researchers and help raise the bar for research quality across the team
• Contribute to external presence through papers, talks, and recruiting

What You Bring

• Strong track record of ML research, ideally with publications in top-tier venues (e.g., NeurIPS, ICML, ICLR, ACL, EMNLP, AAAI, etc.)
• Proven ability to distinguish high-quality vs low-quality research, especially in fast-moving areas like LLMs
• Deep understanding of LLM failure modes, particularly hallucinations, and how to evaluate and mitigate them
• Experience designing rigorous evaluation frameworks and building high-quality test datasets
• Strong intuition for dataset quality, bias, and benchmark design (data-centric AI mindset)
• Hands-on experience training large-scale deep learning models (multi-GPU / distributed systems)
• Deep understanding of modern neural architectures (transformers, SSMs, encoder/decoder models, etc.)
• Strong programming skills in Python and ML frameworks such as PyTorch or JAX
• Experience deploying ML models into production systems and monitoring their performance
• Clear and proactive communicator, able to explain complex ideas and critique work effectively

Nice To Haves

• Experience with inference optimization techniques (e.g., vLLM, KV caching, s