Director, Data Science, Foundation Model AI

Merck Merck · Pharma · MA

Director of foundational AI at Merck, leading a team of ML researchers and engineers to develop and train large multi-modal foundation models and bespoke methods for biological data analysis. The role involves providing technical guidance, interpreting results, and publishing research findings, with a focus on advancing therapeutic strategies. Requires a PhD and significant experience in ML/AI, team leadership, and training large models.

What you'd actually do

  1. Oversee the training of large multi-modal foundation models.
  2. Work with data, including various omics, imaging, and text modalities.
  3. Interpret and critically analyze results from machine learning and AI models.
  4. Provide deep technical guidance.
  5. Coach and advance the team of machine learning researchers.

Skills

Required

  • Python
  • PyTorch
  • classical machine learning
  • probabilistic models
  • causal analysis
  • deep learning frameworks
  • team leadership
  • training large models
  • multi-modal foundation models
  • post-training foundation models
  • parameter-efficient fine-tuning
  • preference optimization

Nice to have

  • transformer-based models
  • generative modeling
  • diffusion modeling
  • flow matching
  • reinforcement learning
  • biological data
  • biological foundation models

What the JD emphasized

  • PhD in Computer Science, Statistics, Physics, Engineering, Mathematics, Data Science, AI/Machine Learning, Computational Biology, Bioinformatics, Computational Biology, or related STEM field and 7+ years of full-time experience or an MS and 10+ years of experience
  • Demonstrated world-class expertise in at least one sub-area of machine learning or AI, as shown by publications in NeurIPS, ICML, ICLR, AISTATS, or equivalent venues, and/or open-source projects
  • Experience training large models on multi-node, multi-GPU environments
  • Experience designing novel architectures for multi-modal foundation models
  • Experience in post-training foundation models, including familiarity with parameter-efficient fine-tuning, post-hoc interpretability, and preference optimization

Other signals

  • leading a team of machine learning researchers and engineers
  • overseeing both large foundation model development and the development of bespoke methods
  • training large multi-modal foundation models
  • post-training foundation models