Sr. Applied Scientist , Grocery, Retail & In-store Experience (graise)

Amazon Amazon · Big Tech · IN, KA, Bengaluru · Applied Science

Senior Applied Scientist role focused on computer vision and machine learning for Amazon's grocery ecosystem. The role involves owning the science strategy, developing novel approaches to perception challenges, defining evaluation frameworks, leading technical design for production systems, and influencing technical direction across teams. It requires driving solutions from research to production at scale, with a focus on multimodal data and real-world applications in a consumer-facing domain.

What you'd actually do

  1. Own the end-to-end science strategy for computer vision and machine learning solutions in the grocery domain, navigating ambiguity to identify the highest-impact opportunities
  2. Develop novel approaches to complex, unsolved perception and identification challenges where off-the-shelf methods are insufficient; publish findings internally or externally to advance the state of the art
  3. Define the evaluation framework and success criteria for model performance, establishing metrics that connect scientific outcomes to measurable business impact and using these to influence roadmap prioritization
  4. Lead cross-functional technical design with engineering, product, and operations partners, driving architecture decisions for model serving, data pipelines, and system reliability at scale rather than solely handing off models for productionization
  5. Identify and resolve ambiguous, cross-team technical dependencies (e.g., upstream data quality, annotation infrastructure, model interoperability) that block progress across multiple workstreams; propose and drive solutions proactively

Skills

Required

  • 5+ years of building machine learning models for business application experience
  • PhD, or Master's degree
  • Experience programming in Java, C++, Python or related language
  • Experience with neural deep learning methods and machine learning

Nice to have

  • Experience leading end-to-end science efforts from problem formulation through production launch
  • Hands-on experience building, training, and deploying computer vision models in production systems operating at scale on real-world, noisy data
  • Experience building and improving continuous model training pipelines, including human-in-the-loop annotation workflows, active learning, and data flywheel strategies that compound model quality over time
  • Demonstrated ability to work with multimodal data (images, video, sensor signals, text/catalog metadata) and design systems that fuse heterogeneous inputs for robust inference
  • Familiarity with retail, logistics, or physical-world perception domains where environmental variability (lighting, angles, occlusion, sensor diversity) makes controlled-lab performance an unreliable predictor of real-world accuracy

What the JD emphasized

  • own the science strategy
  • drive solutions from research through production at scale
  • lead cross-functional technical decisions
  • define the intelligence layer powering millions of grocery shopping experiences
  • develop novel approaches to complex, unsolved perception and identification challenges
  • define the evaluation framework and success criteria for model performance
  • lead cross-functional technical design with engineering, product, and operations partners
  • driving architecture decisions for model serving, data pipelines, and system reliability at scale
  • identify and resolve ambiguous, cross-team technical dependencies
  • influence technical direction beyond the immediate team
  • mentor scientists, raise the bar in hiring, establish best practices for experimentation and model development
  • represent the team's science strategy to senior leadership
  • communicate complex technical trade-offs and recommendations to VP-level stakeholders
  • own the technical strategy for how computer vision and multimodal learning come together to solve perception problems in grocery stores
  • diagnose a surprising failure mode from overnight experiments and decide whether to pivot your approach entirely
  • co-architect a serving system with engineers while defining confidence thresholds and graceful degradation paths
  • present a precision-recall trade-off to senior leaders in terms that shape launch decisions and investment priorities
  • unblock a cross-team dependency on annotation infrastructure
  • mentoring junior scientists and carving out time for deep technical work on problems the team hasn't cracked
  • build foundational AI and machine learning systems that power Amazon's in-store grocery technologies
  • develop domain-specific models that solve uniquely complex challenges in grocery
  • build machine learning models for business application experience
  • building, training, and deploying computer vision models in production systems operating at scale on real-world, noisy data
  • building and improving continuous model training pipelines
  • design systems that fuse heterogeneous inputs for robust inference
  • familiarity with retail, logistics, or physical-world perception domains

Other signals

  • own the science strategy
  • drive solutions from research through production at scale
  • lead cross-functional technical decisions
  • define the intelligence layer powering millions of grocery shopping experiences
  • develop novel approaches to complex, unsolved perception and identification challenges
  • define the evaluation framework and success criteria for model performance
  • lead cross-functional technical design with engineering, product, and operations partners
  • driving architecture decisions for model serving, data pipelines, and system reliability at scale
  • identify and resolve ambiguous, cross-team technical dependencies
  • influence technical direction beyond the immediate team
  • mentor scientists, raise the bar in hiring, establish best practices for experimentation and model development
  • represent the team's science strategy to senior leadership
  • communicate complex technical trade-offs and recommendations to VP-level stakeholders
  • own the technical strategy for how computer vision and multimodal learning come together to solve perception problems in grocery stores
  • diagnose a surprising failure mode from overnight experiments and decide whether to pivot your approach entirely
  • co-architect a serving system with engineers while defining confidence thresholds and graceful degradation paths
  • present a precision-recall trade-off to senior leaders in terms that shape launch decisions and investment priorities
  • unblock a cross-team dependency on annotation infrastructure
  • mentoring junior scientists and carving out time for deep technical work on problems the team hasn't cracked
  • build foundational AI and machine learning systems that power Amazon's in-store grocery technologies
  • develop domain-specific models that solve uniquely complex challenges in grocery
  • build machine learning models for business application experience
  • building, training, and deploying computer vision models in production systems operating at scale on real-world, noisy data
  • building and improving continuous model training pipelines
  • design systems that fuse heterogeneous inputs for robust inference
  • familiarity with retail, logistics, or physical-world perception domains