Principal Applied Scientist, Agentforce Operations

Salesforce Salesforce · Enterprise · Seattle Metro -, New York - New York City Metro -, New York - New York, California - San Francisco Metro -, California - San Francisco, WA · Remote

Salesforce is seeking a Principal Applied Scientist for their Agentforce Operations team. The role focuses on designing, implementing, and training novel deep learning models for enterprise data, prototyping new architectures, and turning research into production-ready features. The ideal candidate has a Master's or PhD, experience with transformer architectures and retrieval mechanisms, a track record of publications, and strong Python skills with experience in distributed training frameworks.

What you'd actually do

  1. Design, implement, and train novel deep learning models on large-scale GPU clusters.
  2. Prototype new architectures and algorithms for enterprise data, such as tabular, relational, and graph-structured data.
  3. Stay current with AI research, apply relevant breakthroughs, and advise on internal AI strategy — wearing multiple hats as part of a startup-style team that values experimental rigor and shipping well-engineered systems.
  4. Partner with engineering, product, and design teams to turn research into functional, production-ready features that create immediate, tangible customer value.

Skills

Required

  • Master's or PhD in Computer Science, Mathematics, or a highly quantitative field with an AI/ML research focus
  • Experience developing, deploying, and evaluating machine learning models in production or top-tier research environments
  • Deep understanding of transformer architectures and advanced retrieval mechanisms
  • Designing novel, non-standard AI solutions
  • Introducing core architectural changes
  • Applying foundational mathematical disciplines (stochastic modeling, graph theory, optimization algorithms)
  • Track record of peer-reviewed publications
  • Production-grade Python skills
  • Experience with automatic differentiation frameworks (PyTorch, JAX, or TensorFlow)
  • Experience training large-scale models across distributed GPU clusters using frameworks like DeepSpeed or TorchTitan

Nice to have

  • Experience building production-grade ML pipelines using tools like Kubeflow, Airflow, or MLflow for tracking, versioning, and automated retraining
  • Experience building robust, distributed data pipelines for cleaning and analyzing massive datasets using tools like PySpark

What the JD emphasized

  • novel deep learning models
  • prototype new architectures
  • enterprise data
  • AI research
  • experimental rigor
  • production-ready features
  • Master's or PhD
  • AI/ML research focus
  • transformer architectures
  • advanced retrieval mechanisms
  • novel, non-standard AI solutions
  • foundational mathematical disciplines
  • stochastic modeling
  • graph theory
  • optimization algorithms
  • track record of peer-reviewed publications
  • production-grade Python skills
  • automatic differentiation frameworks
  • PyTorch, JAX, or TensorFlow
  • training large-scale models
  • distributed GPU clusters
  • DeepSpeed or TorchTitan

Other signals

  • novel deep learning models
  • prototype new architectures
  • turn research into functional, production-ready features
  • production-grade Python skills
  • training large-scale models