Amazon Science homepage
Visit the Amazon Science blog to learn about the latest research and papers from Amazon scientists across artificial intelligence, machine learning, computer vision, quantum, and more.
- Posts
- 105 from 10 Apr 2018 to 29 Oct 2026
- Posting
- weekly (36 in the last year)
- Kind
- company blog
- Feed
- https://www.amazon.science/index.rss
2026 35
- Speeding up language models for travel applications
- Better speech synthesis for audiobooks
- The science behind TorchNeuron, Amazon’s new PyTorch backend for AWS Trainium chips
- Strengthening AI safety evaluation through independent third-party review
- Can we change an AI model’s architecture without retraining it from scratch?
- Graph-centric agentic intelligence
- Latest news
- Recipients
- A kernel-centric path to real-time video generation on Trainium
- Amazon launches research initiative with Stanford University to advance AI and science
- Advancing AI for biology: Teaching models to design and characterize antibodies
- Why don’t machine learning research agents overfit?
- Developing provably correct Rust code with Verus
- When LLM judges agree, should we believe them?
- SOP-Bench: A new benchmark for evaluating AI agents on real business procedures
- A decade of mathematical certainty: Reflections on the Automated Reasoning Group
- AWS Trainium Frontier competition: Co-design models and kernels on purpose-built AI chips
- 34 Amazon Research Awards Build on Trainium recipients announced
- How controllers from industrial machinery can coordinate multitask machine learning
- A new benchmark for evaluating patient-facing health AI agents
- Amazon is investing in the Lean Focused Research Organization
- Amazon and University of Michigan give robots a sense of touch
- Capturing token IDs during agentic interactions for better reinforcement learning
- How Amazon tracks carbon intensity across its operations
- The fuel of the future is already here: Why TRISO matters
- EC2’s formally verified “isolation engine” provides mathematical assurance of virtual-machine isolation
- Graviton5’s improved design increases speed and energy efficiency — beyond Moore’s law
- Real-world grounding in agentic AI
- Bridging intent and execution in agentic systems
- Ground truth is a process, not a dataset
- How flat is replacing fat in AWS data center networks
- Introducing the perception agent harness with annotation and verification
- Diverse reasoning traces teach LLMs to make better decisions
- Making LLMs faster without sacrificing accuracy
- AWS and Hopkins Engineering announce database for AI/ML antibody design
2025 2
2023 1
2019 29
- Amazon Mentors Help UMass Graduate Students Make Concrete Advances on Vital Machine Learning Problems
- How to do fast, accurate multi-category classification
- Active learning: Algorithmically selecting training data to improve Alexa’s natural-language understanding
- Adapting Alexa to regional language variations
- Teaching Alexa to follow conversations
- Amazon Unveils Novel Alexa Dialog Modeling for Natural, Cross-Skill Conversations
- Using adversarial training to recognize speakers’ emotions
- Should Alexa read “2/3” as “two-thirds” or “February Third”?: The science of text normalization
- Training a Machine Learning Model in English Improves Its Performance in Japanese
- How we add new skills to Alexa’s name-free skill selector
- “Alexa, Turn Down the Lights and Play Music”: The Science of Handling Compound Requests
- Training Speech Synthesizers on Data from Multiple Speakers Improves Performance, Stability
- Using wake word acoustics to filter out background speech improves speech recognition by 15%
- Two new papers discuss how Alexa recognizes sounds
- Signal processor improves Echo’s bass response, loudness, and speech recognition accuracy
- Cross-lingual transfer learning for bootstrapping AI systems reduces new-language data requirements
- New speech recognition experiments demonstrate how machine learning can scale
- Joint training on speech signal isolation and speech recognition improves performance
- Audio watermarking algorithm is first to solve "second-screen problem" in real time
- Adversarial training produces synthetic data for machine learning
- Machine-labeled data + artificial noise = better speech recognition
- To correct imbalances in training data, don’t oversample: Cluster
- Innovations from the 2018 Alexa Prize
- AI tools let Alexa Prize participants focus on science
- Why Alexa won't wake up when she hears her name in Amazon's Super Bowl ad
- Updating neural networks to recognize new categories, with minimal retraining
- More-efficient “kernel methods” dramatically reduce training time for natural-language-understanding systems
- Leveraging unannotated data to bootstrap Alexa functions more quickly
- New method for compressing neural networks better preserves accuracy
2018 36
- How Alexa may learn to retrieve stored "memories"
- How Alexa knows “peanut butter” is one shopping-list item, not two
- With New Data Representation Scheme, Alexa Can Better Match Skills to Customer Requests
- New Approach to Language Modeling Reduces Speech Recognition Errors by Up to 15%
- Distributed “Re-Ranker” ensures that Alexa improvements reach customers ASAP
- The role of context in redefining human-computer interaction
- Context-aware deep-learning method boosts Alexa dialogue system’s ability to recognize conversation topics by 35%
- Varying speaking styles with neural text-to-speech
- Reducing Customer Friction through Skill Selection
- Amazon helps launch workshop on automatic fact verification
- How an Echo device could locate snaps, claps, and taps
- Identifying sounds in audio streams
- How Voice and Graphics Working Together Enhance the Alexa Experience
- Whisper to Alexa, and She’ll Whisper Back
- Learning to Recognize the Irrelevant
- How Alexa Is Learning to Ignore TV, Radio, and Other Media Players
- Alexa at Interspeech 2018: How interaction histories can improve speech understanding
- How Alexa is learning to converse more naturally
- 3 questions about Interspeech 2018 with Björn Hoffmeister
- Alexa, do I need to use your wake word? How about now?
- Public release of fact-checking dataset quickly begins to pay dividends
- Shrinking machine learning models for offline use
- Automatic transliteration can help Alexa find data across language barriers
- Contextual Clues Can Help Improve Alexa’s Speech Recognizers
- How Alexa can use song-playback duration to learn customers’ preferences
- 2018 Amazon Research Awards CFP launch announcement
- HypRank: How Alexa determines what skill can best meet a customer’s need
- The Scalable Neural Architecture behind Alexa’s Ability to Select Skills
- New way to annotate training data should enable more sophisticated Alexa interactions
- Machine translation accelerates how Alexa learns new languages
- Amazon scientists use transfer learning to accelerate development of new Alexa capabilities
- Amazon Scientist Outlines Multilayer System For Smart Speaker Echo Cancellation And Voice Enhancement
- Amazon and University of Sheffield researchers make large-scale fact extraction and verification dataset publicly available
- Making Alexa more friction-free
- Alexa scientists present two new techniques that improve wake word performance
- Alexa scientists address challenges of end-pointing