Applied Scientist Ii, Trustworthy Shopping Experience (tse)

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

Applied Scientist II on Amazon's Trustworthy Shopping Experience (TSE) team, focusing on building and productionizing generative AI solutions for automating complex manual investigation processes at scale. The role involves designing and building agentic AI systems with multi-step reasoning, autonomous task execution, and multimodal intelligence, leveraging techniques like SFT, RFT, and few-shot approaches. The scientist will also work on prompt optimization, novel Finetuned transformer architectures, and identifying business problems to apply state-of-the-art LLM workflows. The role offers end-to-end ownership from research to production deployment, with a focus on impacting cost-of-serving customers while maintaining trust and safety.

What you'd actually do

  1. Design and build expertise deep agentic AI systems with multi-step reasoning, autonomous task execution, and multimodal intelligence with capabilities to handle feedback with long term as well as short term memory mechanisms.
  2. Productionize large scale models built on top of SFT (Supervised Finetuning) and RFT (Reinforced fine tuning) approaches (GRPO with RLVR, Process/Outcome Reward Models), few shot approaches (Contrastive, Prototypical) based on multimodal datasets
  3. Build novel production ready Finetuned transformer architectures (using LORA/Q-LORA/LLM-JEPA etc) and conventional supervised & unsupervised ML solutions to aid the multiple potential automation requirements
  4. Identify customer and business problems at project level; invent or extend state-of-the-art approaches for complex LLM workflows involving unstructured text, documents, images, and relational data
  5. Prototype rapidly, iterate based on feedback, and deliver components at SDE 1+ level that integrate directly into production-scale systems

Skills

Required

  • building models for business application experience
  • PhD, or Master's degree and 3+ years of CS, CE, ML or related field experience
  • Experience in patents or publications at top-tier peer-reviewed conferences or journals
  • Experience programming in Java, C++, Python or related language
  • Experience in any of the following areas: algorithms and data structures, parsing, numerical optimization, data mining, parallel and distributed computing, high-performance computing

Nice to have

  • Experience using Unix/Linux
  • Experience in professional software development

What the JD emphasized

  • lead the development of next Gen AI solutions
  • automate complex manual investigation processes at Amazon scale
  • building systems that reason and act autonomously
  • adapt rapidly from limited examples
  • seamlessly connect visual and textual understanding
  • compress complex model capabilities into efficient, deployable systems
  • end-to-end ownership—from initial research and proof-of-concept through production deployment
  • serving hundreds of millions of customers within months, not years
  • Design and build expertise deep agentic AI systems with multi-step reasoning, autonomous task execution, and multimodal intelligence with capabilities to handle feedback with long term as well as short term memory mechanisms.
  • Productionize large scale models built on top of SFT (Supervised Finetuning) and RFT (Reinforced fine tuning) approaches (GRPO with RLVR, Process/Outcome Reward Models), few shot approaches (Contrastive, Prototypical) based on multimodal datasets
  • Enhance on existing Automatic prompt optimization techniques (GEPA & beyond) towards agentic optimization given the ground truth datasets to improve agentic planning.
  • Build novel production ready Finetuned transformer architectures (using LORA/Q-LORA/LLM-JEPA etc) and conventional supervised & unsupervised ML solutions to aid the multiple potential automation requirements
  • Identify customer and business problems at project level; invent or extend state-of-the-art approaches for complex LLM workflows involving unstructured text, documents, images, and relational data
  • Prototype rapidly, iterate based on feedback, and deliver components at SDE 1+ level that integrate directly into production-scale systems
  • Engineer efficient systems balancing model capability, deployment cost, and resource usage; write significant code demonstrating technical excellence and maintainability
  • Scrutinize algorithm and software performance for improvements; resolve root causes leaving systems more maintainable
  • Contribute to tactical and strategic planning—team goals, priorities, and roadmaps—while providing architectural guidance for AI systems
  • Participate in engineering best practices with rigorous peer reviews; communicate design decisions clearly and participate in science reviews
  • Train new teammates & interns on component construction and integration; mentor less experienced scientists and participate in hiring processes

Other signals

  • building intelligent systems that reason, act, and learn
  • automate complex manual investigation processes
  • end-to-end ownership—from initial research and proof-of-concept through production deployment
  • innovations will deliver significant impact to cost-of-serving customers