Senior Applied Scientist - Edge AI Models

About Amazon Devices

Amazon Devices is a pioneering R&D division that designs and develops cutting-edge products including Kindle readers, Fire Tablets, Fire TV, Health & Wellness devices, Amazon Echo, and Astro robots. This role presents a fantastic chance to contribute to innovation by advancing generative AI techniques optimized for edge computing in consumer devices.

Key Responsibilities

• Optimize generative AI models for edge deployment through quantization, pruning, distillation, and fine-tuning methods.
• Acquire in-depth knowledge of Amazon's Neural Edge Engine architectures and devise novel optimization approaches.
• Analyze deep learning workloads and strategize their efficient implementation on Amazon's specialized Neural Edge Engine.
• Apply advanced principles from information theory, scientific computing, deep learning theory, and non-equilibrium thermodynamics to solution development.
• Design and train custom generative AI models that surpass current state-of-the-art benchmarks and set foundations for production readiness.
• Collaborate cross-functionally with compiler engineers, fellow applied scientists, hardware architects, and product teams to craft superior ML-driven solutions for hardware devices.
• Represent Amazon in the wider research community by publishing open source work and presenting at premier conferences such as NeurIPS, ICLR, and MLSys.

Mandatory Qualifications

• Minimum 3 years of experience developing machine learning models for practical business use cases.
• PhD degree or Master's with at least four years of applied research professional experience.
• Proficient programming skills in Java, C++, Python, or comparable languages.
• Proven expertise in neural deep learning techniques and general machine learning methods.
• Prior involvement in the full lifecycle management of ML models, from creation to production deployment.

Preferred Skills

• Knowledge and preferably practical experience with advanced inference optimization techniques such as Mixture-of-Experts (MoE) architectures and Diffusion Models tailored for language generation.

Additional Information

Amazon embraces inclusivity and supports applicants with disabilities by providing accommodations throughout the recruitment process. Candidates can request assistance during interviews and onboarding. For those in regions not listed, coordination with the recruiting partner is advised.

Work Location: ADCI - Bengaluru 14 SEZ