Lead Principal Applied Scientist - Robotics

Oracle Oracle · Enterprise · United States

Lead Principal Applied Scientist for Robotics, Perception, and Embodied AI at Oracle. This role focuses on defining, prototyping, and delivering AI capabilities for commercial robotic systems. Responsibilities include applied research in multi-sensor fusion, perception, multimodal reasoning, and reinforcement learning, with an emphasis on full-stack delivery from data to production deployment. The role requires strong collaboration across science, engineering, product, and hardware teams.

What you'd actually do

  1. Partner with science, engineering, product, and hardware leaders to identify strategic product needs where robotics, perception, and embodied AI can create measurable customer and business impact.
  2. Lead the design and execution of research programs and POCs for sensors, multi-sensor fusion, real-time signal processing, and perception systems.
  3. Lead full-stack execution across experimentation, data pipelines, model development, evaluation, deployment integration, and production monitoring.
  4. Demonstrate thought leadership in at least one business-critical area such as robot perception, multimodal systems, sensor fusion, reinforcement learning, or embodied AI.
  5. Mentor and guide scientists and engineers, raising the bar for applied research rigor, experimentation quality, and production readiness.

Skills

Required

  • machine learning
  • artificial intelligence
  • computer vision
  • perception
  • robotics
  • sensor fusion
  • real-time signal processing
  • object detection
  • tracking
  • activity recognition
  • scene understanding
  • localization
  • state estimation
  • multimodal AI systems
  • reinforcement learning
  • planning
  • sequential decision-making
  • robotics control interfaces
  • action-conditioned model behavior
  • Python
  • C++
  • modern ML frameworks
  • production-oriented software practices
  • APIs
  • SDKs
  • commercially viable robotics platforms

Nice to have

  • data collection
  • experimentation
  • model architectures
  • evaluation criteria
  • production milestones
  • perception accuracy
  • latency
  • robustness
  • safety
  • reliability
  • operational feedback metrics
  • modeling approaches
  • evaluation techniques
  • data collection strategies
  • risk/reward tradeoffs
  • scene understanding
  • multimodal reasoning
  • action-conditioned planning
  • decision-making
  • robot behaviors
  • dataset strategy
  • data quality criteria
  • labeling approaches
  • simulation
  • synthetic data
  • evaluation protocols
  • model training
  • fine-tuning
  • optimization
  • inference design
  • compute/latency tradeoffs
  • real-time deployment
  • near-real-time deployment
  • data pipelines
  • deployment integration
  • production monitoring
  • machine learning engineering
  • hardware teams
  • product stakeholders
  • code review
  • documentation
  • testing
  • delivery readiness
  • root-cause analysis
  • thought leadership
  • robot perception
  • sensor fusion
  • embodied AI
  • patents
  • white papers
  • design documents
  • demos
  • conference-quality publications
  • mentoring
  • applied research rigor
  • experimentation quality
  • production readiness
  • external research groups
  • academic partners
  • commercial robotics ecosystem partners
  • ambiguous idea to prototype
  • cross-functional collaboration
  • product urgency
  • systems constraints
  • customer impact
  • executive-level communication
  • technical audiences
  • non-technical audiences

What the JD emphasized

  • hands-on execution
  • full-stack delivery
  • hands-on orientation
  • full-stack AI workflows

Other signals

  • delivering AI capabilities for commercially viable robotic systems
  • end-to-end solution development from data collection and experimentation through production deployment
  • applied research in multi-sensor fusion, real-time signals, perception, multimodal reasoning, reinforcement learning, and action-conditioned planning
  • hands-on execution and full-stack delivery