(usa) Staff, Data Scientist

Walmart Walmart · Retail · Sunnyvale, CA

Staff Data Scientist at Walmart focused on designing, implementing, and scaling supervised fine-tuning (SFT) pipelines for foundation models, including instruction following and ranking tasks. The role involves optimizing training hyperparameters, managing dataset mixtures, and applying preference optimization techniques like RLHF. The team leverages large-scale user interactions to build models for e-commerce applications such as search, advertising, recommendation, and personalization.

What you'd actually do

  1. Design, implement, and scale high-throughput SFT pipelines to train foundation models for instruction following, ranking, and related tasks.
  2. Manage high-quality dataset mixtures and optimize training hyperparameters, including learning rate schedules and other training configurations.
  3. Drive model alignment toward helpfulness, accuracy, and safety through preference optimization techniques and reinforcement learning approaches.
  4. Collaborate with cross-functional teams to deploy, monitor, and improve machine learning models in production environments.
  5. Mentor junior team members on analytical methodologies, coding best practices, machine learning techniques, and data storytelling.

Skills

Required

  • Proven expertise in data science, including predictive analytics, machine learning algorithms, statistical modeling, and hypothesis testing.
  • Strong programming skills in Python and experience with data visualization tools such as Matplotlib, Plotly, Tableau, or Power BI.
  • Hands-on experience designing, training, and deploying machine learning models, with a focus on transformer-based architectures and large language models.
  • Experience optimizing large-scale training pipelines, including hyperparameter tuning and implementing high-throughput workflows using frameworks such as PyTorch.
  • Experience applying reinforcement learning methods (e.g., RLHF) for preference optimization, model alignment, and safety-focused applications.
  • Demonstrated ability to translate complex business requirements into analytical models and actionable business insights.
  • Excellent verbal and written communication skills, with the ability to present technical findings effectively to both technical and non-technical stakeholders.

What the JD emphasized

  • high-throughput SFT pipelines
  • foundation models
  • transformer-based architectures
  • large language models
  • optimizing large-scale training pipelines
  • reinforcement learning methods (e.g., RLHF)

Other signals

  • foundation models
  • SFT pipelines
  • transformer-based architectures
  • large language models
  • RLHF