Alexa speech science developments at Interspeech 2022

Research from Alexa Speech covers a range of topics related to end-to-end neural speech recognition and fairness.

Interspeech, the world’s largest and most comprehensive conference on the science and technology of spoken-language processing, took place this week in Incheon, Korea, with Amazon as a platinum sponsor. Amazon Science asked three of Alexa AI’s leading scientists — in the fields of speech, spoken-language-understanding, and text-to-speech — to highlight some of Amazon’s contributions to the conference.

Related content
Methods for learning from noisy data, using phonetic embeddings to improve entity resolution, and quantization-aware training are a few of the highlights.

In this installment, senior principal scientist Andreas Stolcke selects papers from Alexa AI’s speech science organization, focusing on two overarching themes in recent research on speech-enabled AI: end-to-end neural speech recognition and fairness.

End-to-end neural speech recognition

Traditionally, speech recognition systems have included components specialized for different aspects of linguistic knowledge: acoustic models to capture the correspondence between speech sounds and acoustic waveforms (phonetics), pronunciation models to map those sounds to words, and language models (LMs) to capture higher-order properties such as syntax, semantics, and dialogue context.

All these models are trained on separate data and combined using graph and search algorithms, to infer the most probable sequence of words corresponding to acoustic input. The latest versions of these systems employ neural networks for individual components, typically in the acoustic and language models, while still relying on non-neural methods for model integration; they are therefore known as “hybrid” automatic-speech-recognition (ASR) systems.

While the hybrid ASR approach is structured and modular, it also makes it hard to model the ways in which acoustic, phonetic, and word-level representations interact and to optimize the recognition system end to end. For these reasons, much recent research in ASR has focused on so-called end-to-end or all-neural recognition systems, which infer a sequence of words directly from acoustic inputs.

Related content
Innovative training methods and model compression techniques combine with clever engineering to keep speech processing local.

End-to-end ASR systems use deep multilayered neural architectures that can be optimized end to end for recognition accuracy. While they do require large amounts of data and computation for training, once trained, they offer a simplified computational architecture for inference, as well as superior performance.

Alexa’s ASR employs end-to-end as its core algorithm, both in the cloud and on-device. Across the industry and in academic research, end-to-end architectures are still being improved to achieve better accuracy, to require less computation and/or latency, or to mitigate the lack of modularity that makes it challenging to inject external (e.g., domain-specific) knowledge at run time.

Alexa AI papers at Interspeech address several open problems in end-to-end ASR, and we summarize a few of those papers here.

In “ConvRNN-T: Convolutional augmented recurrent neural network transducers for streaming speech recognition”, Martin Radfar and coauthors propose a new variant of the popular recurrent-neural-network-transducer (RNN-T) end-to-neural architecture. One of their goals is to preserve the property of causal processing, meaning that the model output depends only on past and current (but not future) inputs, which enables streaming ASR. At the same time, they want to improve the model’s ability to capture long-term contextual information.

ConvRNN.png
A high-level block diagram of ConvRNN-T.

To achieve both goals, they augment the vanilla RNN-T with two distinct convolutional (CNN) front ends: a standard one for encoding correlations localized in time and a novel “global CNN” encoder that is designed to capture long-term correlations by summarizing activations over the entire utterance up to the current time step (while processing utterances incrementally through time).

The authors show that the resulting ConvRNN-T gives superior accuracy compared to other proposed neural streaming ASR architectures, such as the basic RNN-T, Conformer, and ContextNet.

Another concern with end-to-end ASR models is computational efficiency, especially since the unified neural architecture makes these models very attractive for on-device deployment, where compute cycles and (for mobile devices) power are at a premium.

In their paper “Compute cost amortized Transformer for streaming ASR”, Yi Xie and colleagues exploit the intuitive observation that the amount of computation a model performs should vary as a function of the difficulty of the task; for instance, input in which noise or an accent causes ambiguity may require more computation than a clean input with a mainstream accent. (We may think of this as the ASR model “thinking harder” in places where the words are more difficult to discern.)

Related content
A new approach to determining the “channel configuration” of convolutional neural nets improves accuracy while maintaining runtime efficiency.

The researchers achieve this with a very elegant method that leverages the integrated neural structure of the model. Their starting point is a Transformer-based ASR system, consisting of multiple stacked layers of multiheaded self-attention (MHA) and feed-forward neural blocks. In addition, they train “arbitrator” networks that look at the acoustic input (and, optionally, also at intermediate block outputs) to toggle individual components on or off.

Because these component blocks have “skip connections” that combine their outputs with the outputs of earlier layers, they are effectively optional for the overall computation to proceed. A block that is toggled off for a given input frame saves all the computation normally carried out by that block, producing a zero vector output. The following diagram shows the structure of both the elementary Transformer building block and the arbitrator that controls it:

Arbitrator:Transformer backbone.png
Illustration of the arbitrator and Transformer backbone of each block. The lightweight arbitrator toggles whether to evaluate subcomponents during the forward pass.

The arbitrator networks themselves are small enough that they do not contribute significant additional computation. What makes this scheme workable and effective, however, is that both the Transformer assemblies and the arbitrators that control them can be trained jointly, with dual goals: to perform accurate ASR and to minimize the overall amount of computation. The latter is achieved by adding a term to the training objective function that rewards reducing computation. Dialing a hyperparameter up or down selects the desired balance between accuracy and computation.

Related content
Branching encoder networks make operation more efficient, while “neural diffing” reduces bandwidth requirements for model updates.

The authors show that their method can achieve a 60% reduction in computation with only a minor (3%) increase in ASR error. Their cost-amortized Transformer proves much more effective than a benchmark method that constrains the model to attend only to sliding windows over the input, which yields only 13% savings and an error increase of almost three times as much.

Finally, in this short review of end-to-end neural ASR advances, we look at ways to recognize speech from more than one speaker, while keeping track of who said what (also known as speaker-attributed ASR).

This has traditionally been done with modular systems that perform ASR and, separately, perform speaker diarization, i.e., labeling stretches of audio according to who is speaking. However, here, too, neural models have recently brought advances and simplification, by integrating these two tasks in a single end-to-end neural model.

In their paper “Separator-transducer-segmenter: Streaming recognition and segmentation of multi-party speech”, Ilya Sklyar and colleagues not only integrate ASR and segmentation-by-speaker but do so while processing inputs incrementally. Streaming multispeaker ASR with low latency is a key technology to enable voice assistants to interact with customers in collaborative settings. Sklyar’s system does this with a generalization of the RNN-T architecture that keeps track of turn-taking between multiple speakers, up to two of whom can be active simultaneously. The researchers’ separator-transducer-segmenter model is depicted below:

Separator-transducer-segmenter.png
Separator-transducer-segmenter. The tokens <sot> and <eot> represent the start of turn and end of turn. Model blocks with the same color have tied parameters, and transcripts in the color-matched boxes belong to the same speaker.

A key element that yields improvements over an earlier approach is the use of dedicated tokens to recognize both starts and ends of speaker turns, for what the authors call “start-pointing” and “end-pointing”. (End-pointing is a standard feature of many interactive ASR systems necessary to predict when a talker is done.) Beyond representing the turn-taking structure in this symbolic way, the model is also penalized during training for taking too long to output these markers, in order to improve the latency and temporal accuracy of the outputs.

Fairness in the performance of speech-enabled AI

The second theme we’d like to highlight, and one that is receiving increasing attention in speech and other areas of AI, is performance fairness: the desire to avert large differences in accuracy across different cohorts of users or on content associated with protected groups. As an example, concerns about this type of fairness gained prominence with demonstrations that certain computer vision algorithms performed poorly for certain skin tones, in part due to underrepresentation in the training data.

Related content
The team’s latest research on privacy-preserving machine learning, federated learning, and bias mitigation.

There’s a similar concern about speech-based AI, with speech properties varying widely as a function of speaker background and environment. A balanced representation in training sets is hard to achieve, since the speakers using commercial products are largely self-selected, and speaker attributes are often unavailable for many reasons, privacy among them. This topic is also the subject of a special session at Interspeech, Inclusive and Fair Speech Technologies, which several Alexa AI scientists are involved in as co-organizers and presenters.

One of the special-session papers, “Reducing geographic disparities in automatic speech recognition via elastic weight consolidation”, by Viet Anh Trinh and colleagues, looks at how geographic location within the U.S. affects ASR accuracy and how models can be adapted to narrow the gap for the worst-performing regions. Here and elsewhere, a two-step approach is used: first, subsets of speakers with higher-than-average error rates are identified; then a mitigation step attempts to improve performance for those cohorts. Trinh et al.’s method identifies the cohorts by partitioning the speakers according to their geographic longitude and latitude, using a decision-tree-like algorithm that maximizes the word-error-rate (WER) differences between resulting regions:

Reducing geographical disparities.png
A map of 126 regions identified by the clustering tree. The color does not indicate a specific word error rate (WER), but regions with the same color do have the same WER.

Next, the regions are ranked by their average WERs; data from the highest-error regions is identified for performance improvement. To achieve that, the researchers use fine-tuning to optimize the model parameters for the targeted regions, while also employing a technique called elastic weight consolidation (EWC) to minimize performance degradation on the remaining regions.

This is important to prevent a phenomenon known as “catastrophic forgetting”, in which neural models degrade substantially on prior training data during fine-tuning. The idea is to quantify the influence that different dimensions of the parameter space have on the overall performance and then avoid large variations along those dimensions when adapting to a data subset. This approach decreases the WER mean, maximum, and variance across regions and even the overall WER (including the regions not fine-tuned on), beating out several baseline methods for model adaptation.

Pranav Dheram et al., in their paper “Toward fairness in speech recognition: Discovery and mitigation of performance disparities”, look at alternative methods for identifying underperforming speaker cohorts. One approach is to use human-defined geographic regions as given by postal (a.k.a. zip) codes, in combination with demographic information from U.S. census data, to partition U.S. geography.

Related content
NSF deputy assistant director Erwin Gianchandani on the challenges addressed by funded projects.

Zip codes are sorted into binary partitions by majority demographic attributes, so as to maximize WER discrepancies. The partition with higher WER is then targeted for mitigations, an approach similar to that adopted in the Trinh et al. paper. However, this approach is imprecise (since it lumps together speakers by zip code) and limited to available demographic data, so it generalizes poorly to other geographies.

Alternatively, Dheram et al. use speech characteristics learned by a neural speaker identification model to group speakers. These “speaker embedding vectors” are clustered, reflecting the intuition that speakers who sound similar will tend to have similar ASR difficulty.

Subsequently, these virtual speaker regions (not individual identities) can be ranked by difficulty and targeted for mitigation, without relying on human labeling, grouping, or self-identification of speakers or attributes. As shown in the table below, the automatic approach identifies a larger gap in ASR accuracy than the “geo-demographic” approach, while at the same time targeting a larger share of speakers for performance mitigation:

Cohort discovery

WER gap (%)

Bottom-cohort share (%)

Geodemographic

Automatic

41.7

65.0

0.8

10.0

The final fairness-themed paper we highlight explores yet another approach to avoiding performance disparities, known as adversarial reweighting (ARW). Instead of relying on explicit partitioning of the input space, this approach assigns continuous weights to the training instances (as a function of input features), with the idea that harder examples get higher weights and thereby exert more influence on the performance optimization.

Related content
Method significantly reduces bias while maintaining comparable performance on machine learning tasks.

Secondly, ARW more tightly interleaves, and iterates, the (now weighted) cohort identification and mitigation steps. Mathematically, this is formalized as a min-max optimization algorithm that alternates between maximizing the error by changing the sample weights (hence “adversarial”) and minimizing the weighted verification error by adjusting the target model parameters.

ARW was designed for group fairness in classification and regression tasks that take individual data points as inputs. “Adversarial reweighting for speaker verification fairness”, by Minho Jin et al., looks at how the concept can be applied to a classification task that depends on pairs of input samples, i.e., checking whether two speech samples come from the same speaker. Solving this problem could help make a voice-based assistant more reliable at personalization and other functions that require knowing who is speaking.

The authors look at several ways to adapt ARW to learning similarity among speaker embeddings. The method that ultimately worked best assigns each pair of input samples an adversarial weight that is the sum of individual sample weights (thereby reducing the dimensionality of the weight prediction). The individual sample weights are also informed by which region of the speaker embedding space a sample falls into (as determined by unsupervised k-means clustering, the same technique used in Dheram et al.’s automatic cohort-identification method).

Computing ARW weights.png
Computing adversarial-reweighting (ARW) weights.

I omit the details, but once the pairwise (PW) adversarial weights are formalized in this way, we can insert them into the loss function for metric learning, which is the basis of training a speaker verification model. Min-max optimization can then take turns training the adversary network that predicts the weights and optimizing the speaker embedding extractor that learns speaker similarity.

On a public speaker verification corpus, the resulting system reduced overall equal-error rate by 7.6%, while also reducing the gap between genders by 17%. It also reduced the error variability across different countries of origin, by nearly 10%. Note that, as in the case of the Trinh et al. ASR fairness paper, fairness mitigation improves both performance disparities and overall accuracy.

This concludes our thematic highlights of Alexa Speech Interspeech papers. Note that Interspeech covers much more than speech and speaker recognition. Please check out companion pieces that feature additional work, drawn from technical areas that are no less essential for a functioning speech-enabled AI assistant: natural-language understanding and speech synthesis.

Research areas

Related content

US, WA, Seattle
Amazon's eCommerce Foundation (eCF) organization provides the core technologies that drive and power Amazon's Stores, Digital, and Other (SDO) businesses. Millions of customer page views and orders per day are enabled by the systems eCF builds from the ground up. CloudTune, within eCF, empowers growth and business agility needs by automatically and efficiently managing AWS capacity and business processes needed to safely meet Amazon's customer demand. CloudTune serves its primary customers, internal software teams, through forecast-driven automation of cost controllership, capacity management, and scaling. We predict expected load, and drive procurement and allocation of AWS capacity for new product launches and high-velocity events like Prime Day and Cyber Monday. CloudTune is looking for a Data Scientist to join our forecasting and optimization team. The team develops sophisticated algorithms that combine machine learning-based demand forecasting with mathematical optimization to solve large-scale capacity planning problems under uncertainty. We work with massive datasets — spanning thousands of availability zone and instance family combinations across global regions — to determine optimal resource allocation strategies that balance infrastructure holding costs against service availability requirements. These models directly inform multi-million dollar capacity investment decisions and drive automated procurement, retention, and release policies across Amazon's compute infrastructure. As a Data Scientist in CloudTune, you will work with other scientists, software engineers, data engineers, and product managers on a variety of important research problems in the areas of stochastic optimization, time series modeling, and operations research. You will work on statistical problems with a high level of ambiguity. You will analyze and process large amounts of data, develop new algorithms and improve existing approaches based on statistical models, machine learning algorithms and big data solutions to automatically scale Amazon’s compute infrastructure, optimizing the balance between availability risk and cost efficiency for all of Amazon businesses. Key job responsibilities - Process and analyze large data sets, mining additional data sources as needed - Build statistical models and drive scalable solutions for multi-year capacity demand forecasting horizons - Create, enhance, and maintain technical documentation, and present to other scientists and business leaders - Process and analyze large-scale operational data to identify patterns and inform model development - Create, enhance, and maintain technical documentation, and present research findings to scientists, engineering teams, and senior leadership
US, CA, Irvine
Do you want to write the algorithms that control next-generation batteries, and shape the future of energy storage? This role sits where electrochemistry, data/ML, and control theory meet. You will develop the physics-informed and data-driven methods that estimate state of charge, health, power capability, and remaining life, and validate them against lab and fleet data until they are ready to ship. Your algorithms will decide how hard a battery can be pushed, when it needs attention, and how long it will last, so they shape the safety, warranty, and economics of the whole product. You will work side by side with cell, systems, and software engineers, and as an early hire you will help define how we develop and validate battery algorithms. This role can be based in Boston, Irvine, Sunnyvale, or Seattle, and involves occasional domestic and international travel, up to 10% of the time. Key job responsibilities • Develop and own algorithms for state of charge, state of health, power capability, and remaining useful life, combining physics-based models with data-driven and estimation methods. • Deliver validated reference implementations, test vectors, and accuracy targets that define what production code must achieve. • Design the test plans and data needed to develop and validate algorithms, working with the cell and testing teams to turn lab and field data into model improvements. • Analyze algorithm performance across the deployed fleet, find where models drift or fail, and drive improvements. • Own the accuracy and robustness of battery algorithms, and sign off that production implementations match the validated reference. Provide algorithm evidence for safety certification. A day in the life Your week moves between data, models, and experiments. One day you might be working out why a health estimate drifts on cells that have aged in the field, and the next deciding which model structure holds accuracy at cold temperatures. You could be designing a test with the cell team to expose a behavior your model misses, or reviewing an implementation with an applied scientist to confirm that what ships matches what you validated. You will regularly share results with software, systems, and cell leaders, and your findings will shape decisions well beyond the algorithm itself. About the team We are building Amazon's energy storage capability, providing power for advanced technologies while contributing to a stable grid. Our product is highly connected, yet secure and resilient by design. We own it end to end, from software, ML, and controls through system design to cell selection and supply, and on to a safety-certified product in the field. The organization operates like a startup inside Amazon: small, fast-moving, and accountable for our decisions. We work across hardware, software, and science, with suppliers and partners worldwide, and with Amazon infrastructure, procurement, legal, and finance. The organization is at an early stage and growing quickly. You will be among the first algorithm hires, so you will help shape how we design, build, certify, and ship, with a direct line of sight to the product and business decisions your work informs. This is a dynamic role that will regularly present new challenges as the product and organization evolve.
US, WA, Redmond
At Amazon, we’re inventing on behalf of customers, and with Amazon Leo, we’re redefining what global connectivity looks like. Our mission is to deliver fast, affordable broadband to unserved and underserved communities around the world through a constellation of low Earth orbit (LEO) satellites. Every system we build helps connect people to education, healthcare, opportunity, and each other. As a Data Scientist, you will be responsible for developing advanced analytics and machine learning solutions for user terminals. You will develop predictive models to proactively identify possible user terminal failures in the field. You will work in a collaborative environment with a multi-disciplinary team, including constellation, RF, antenna, silicon, algorithm, and software engineers. Key job responsibilities As a Data Scientist, you will develop analytic tools for a team developing current and future user terminals. Your responsibilities include: • Develop statistical and analytical tool to enable the regression decision from on-orbit and lab measurement of user terminals • Publish documents and create compelling visualizations and presentations to communicate insights to stakeholders • Create and manage datasets for continued pre-training and supervised fine-tuning of LLMs • Develop scalable visualizations for analysis of user terminal performance • Work closely with constellation, RF, antenna, silicon, algorithm, and software engineers to root-cause the failures using data as the primary tool • Drive consensus on metrics and analysis approaches to support product development strategy Export Control Requirement: Due to applicable export control laws and regulations, candidates must be a U.S. citizen or national, U.S. permanent resident (i.e., current Green Card holder), or lawfully admitted into the U.S. as a refugee or granted asylum. A day in the life As a Data Scientist in the LEO Customer Terminal Team, you will work daily with satellite constellation, algorithm, RF, antenna, silicon, hardware, and software teams in a collaborative environment. Your focus will be using data as an intelligence source to enable design decisions for the team. About the team The LEO Customer Terminal team is responsible for developing both outdoor and indoor devices that enable customers to access internet service via the LEO satellite network. We own the entire process from early prototypes through mass production, including requirements documentation, architecture definition, hardware development, algorithm development, and all integration and verification testing.
US, CA, San Francisco
The Amazon Center for Quantum Computing (CQC) is seeking an Applied Scientist to research and design cryogenic control hardware for large-scale quantum processors, with an emphasis on superconducting digital circuits. Controlling and reading out orders of magnitude more qubits than today's systems demands rethinking where classical control lives and what technology implements it; you will design and model superconducting digital logic (e.g., single-flux-quantum-family circuits) and related cryogenic electronics, evaluate them against the wiring, power, and thermal budgets of a scaled system, and co-design control approaches with the qubit devices they serve. You will carry designs from concept through simulation, layout, design review, and fabrication handoff, and your evaluations will inform which control-electronics directions the organization invests in. Key job responsibilities • Design and simulate superconducting digital circuits and cryogenic control/readout electronics for scalable qubit control • Develop and validate Josephson-junction circuit models, timing and margin analyses, and verification methodology for superconducting digital designs • Quantify system-level trade-offs — wiring count, power dissipation at each cryogenic stage, latency, footprint — across candidate control architectures, and communicate recommendations • Co-design control interfaces with qubit-device colleagues so control electronics and quantum devices are developed as one system • Deliver design artifacts through the organization's design-request and design-review processes, including tapeout-ready layouts and verification reports • Present results internally and, where appropriate, publish in peer-reviewed venues About the team The Enabling Technologies team develops technologies and relationships facilitating scaling, including alternative quantum processor architectures, superconducting qubit device packaging, and cryogenic control device electronics. We work alongside a co-located quantum processor design team focused on current-generation devices; the two teams share design and review processes, and moving ideas between them is an explicit part of our charter.
US, WA, Redmond
Amazon Leo is building a constellation of thousands of low Earth orbit satellites. It will bring fast, reliable internet to customers beyond the reach of existing networks. On the Guidance, Navigation and Control (GNC) team, you will design the algorithms, simulations and onboard autonomy that keep every satellite precisely pointed and stable. Your work lets the payload serve customers on the ground, and it keeps the fleet flying safely. This is satellite GNC at constellation scale. The constellation is launching and growing today, so you will see your work go from concept to orbit, and then fly on many satellites. In this role, you will: - Take your GNC algorithms from concept to orbit, and stay with them through integration, test and on-orbit operations. - Write the production C++ flight code for your own algorithms. The code that flies is the code you write. - Solve hard problems in precise pointing, payload cueing and fleet autonomy. - Work across design, simulation, test and flight operations, not a narrow slice of the lifecycle. **Export Control Requirement:** Due to applicable export control laws and regulations, candidates must be a U.S. citizen or national, U.S. permanent resident (i.e., current Green Card holder), or lawfully admitted into the U.S. as a refugee or granted asylum. Key job responsibilities - Design and analyze algorithms for state estimation, attitude determination and control, precise pointing and payload cueing against demanding performance requirements. - Implement these algorithms in production C++ flight software. Our GNC engineers own their algorithms all the way into flight, and partner with our embedded software engineers to integrate and test them. - Write concepts of operations for each phase of a satellite's life, from launch and orbit raising to service and deorbit, including failures and recovery. Balance what the payload needs against momentum, power, thermal and ground-contact limits. - Develop momentum management strategies with reaction wheels and magnetic torquers. Help select, size and place the GNC sensors and actuators for future satellites, and design the algorithms that use them. These include star trackers, magnetometers, sun sensors, IMUs and GNSS receivers. - Develop onboard autonomy, including Fault Detection, Isolation and Recovery (FDIR), so that the fleet operates safely at scale with minimal ground intervention. - Build and validate satellite and constellation models and simulations at several levels of fidelity, including 6-DOF simulation and Monte Carlo analysis. Verify closed-loop performance in hardware-in-the-loop (HITL) tests. - Support component environmental tests, functional and performance checkout, and subsystem and satellite integration in our labs. - Analyze telemetry from the whole fleet. Build the metrics, dashboards and tools that track pointing performance and payload operations time as the constellation grows, and use them to improve the fleet. - Operate the GNC system on orbit and take part in the team's on-call rotation. Review fleet data, support maneuvers, and resolve anomalies when on-orbit behavior differs from the ground model. - Review designs and code with the team, and document your algorithms, analyses and results clearly. A day in the life This role covers a wide range of problems across the spacecraft and the network. You will derive and prototype estimation and control algorithms, and then turn them into production flight code. You will prove them in 6-DOF and Monte Carlo simulation and in hardware-in-the-loop tests in the lab. You will support satellite integration. When a satellite does not behave as the ground model predicted, you will dig into fleet telemetry, find the cause and drive the fix. You will own each problem end to end, with the focus to go deep on it. GNC is at the center of safe flight, so you will work across the whole program. The payload, power, thermal and operations teams all depend on GNC, and you will balance their needs against the pointing and momentum budgets. You will partner closely with avionics, flight software and embedded software engineers to make your algorithms run reliably on the spacecraft. You will help the mechanical teams shape the next generation of satellite hardware, from reaction wheels to star trackers. As satellites move from design through manufacturing into orbit, you will work with systems engineering and Assembly, Integration and Test. The problems are hard: tight pointing requirements, a large constellation to keep flying safely, and the chance to invent capabilities that do not exist yet. Constant work with other teams makes this a place where you keep learning and growing. About the team You will join the Satellite Attitude Determination and Control team. We own GNC for our satellites end to end: algorithm design, flight software, lab integration, and flight operations of our own satellites on orbit. Our engineers bring deep experience from many satellite programs and other flight vehicles, with real strength in both our mission and the core GNC disciplines. We design, prototype, test, iterate and learn together. Because GNC is central to safe flight, we drive the concepts of operations and many of the system-level analyses on the program. We also have a lot of freedom. We trust our engineers to choose their own tools and approaches for the problems they face, and to own the results. If you want the autonomy to solve hard problems your way, with a team of strong engineers, you will feel at home here.
US, WA, Redmond
Amazon Leo is building a constellation of thousands of low Earth orbit satellites. It will bring fast, reliable internet to customers beyond the reach of existing networks. On the Guidance, Navigation and Control (GNC) team, you will design the algorithms, simulations and onboard autonomy that keep every satellite precisely pointed and stable. Your work lets the payload serve customers on the ground, and it keeps the fleet flying safely. This is satellite GNC at constellation scale. The constellation is launching and growing today, so you will see your work go from concept to orbit, and then fly on many satellites. In this role, you will: - Take your GNC algorithms from concept to orbit, and stay with them through integration, test and on-orbit operations. - Write the production C++ flight code for your own algorithms. The code that flies is the code you write. - Solve hard problems in precise pointing, payload cueing and fleet autonomy. - Work across design, simulation, test and flight operations, not a narrow slice of the lifecycle. **Export Control Requirement:** Due to applicable export control laws and regulations, candidates must be a U.S. citizen or national, U.S. permanent resident (i.e., current Green Card holder), or lawfully admitted into the U.S. as a refugee or granted asylum. Key job responsibilities - Design and analyze algorithms for state estimation, attitude determination and control, precise pointing and payload cueing against demanding performance requirements. - Implement these algorithms in production C++ flight software. Our GNC engineers own their algorithms all the way into flight, and partner with our embedded software engineers to integrate and test them. - Write concepts of operations for each phase of a satellite's life, from launch and orbit raising to service and deorbit, including failures and recovery. Balance what the payload needs against momentum, power, thermal and ground-contact limits. - Develop momentum management strategies with reaction wheels and magnetic torquers. Help select, size and place the GNC sensors and actuators for future satellites, and design the algorithms that use them. These include star trackers, magnetometers, sun sensors, IMUs and GNSS receivers. - Develop onboard autonomy, including Fault Detection, Isolation and Recovery (FDIR), so that the fleet operates safely at scale with minimal ground intervention. - Build and validate satellite and constellation models and simulations at several levels of fidelity, including 6-DOF simulation and Monte Carlo analysis. Verify closed-loop performance in hardware-in-the-loop (HITL) tests. - Support component environmental tests, functional and performance checkout, and subsystem and satellite integration in our labs. - Analyze telemetry from the whole fleet. Build the metrics, dashboards and tools that track pointing performance and payload operations time as the constellation grows, and use them to improve the fleet. - Operate the GNC system on orbit and take part in the team's on-call rotation. Review fleet data, support maneuvers, and resolve anomalies when on-orbit behavior differs from the ground model. - Review designs and code with the team, and document your algorithms, analyses and results clearly. A day in the life This role covers a wide range of problems across the spacecraft and the network. You will derive and prototype estimation and control algorithms, and then turn them into production flight code. You will prove them in 6-DOF and Monte Carlo simulation and in hardware-in-the-loop tests in the lab. You will support satellite integration. When a satellite does not behave as the ground model predicted, you will dig into fleet telemetry, find the cause and drive the fix. You will own each problem end to end, with the focus to go deep on it. GNC is at the center of safe flight, so you will work across the whole program. The payload, power, thermal and operations teams all depend on GNC, and you will balance their needs against the pointing and momentum budgets. You will partner closely with avionics, flight software and embedded software engineers to make your algorithms run reliably on the spacecraft. You will help the mechanical teams shape the next generation of satellite hardware, from reaction wheels to star trackers. As satellites move from design through manufacturing into orbit, you will work with systems engineering and Assembly, Integration and Test. The problems are hard: tight pointing requirements, a large constellation to keep flying safely, and the chance to invent capabilities that do not exist yet. Constant work with other teams makes this a place where you keep learning and growing. About the team You will join the Satellite Attitude Determination and Control team. We own GNC for our satellites end to end: algorithm design, flight software, lab integration, and flight operations of our own satellites on orbit. Our engineers bring deep experience from many satellite programs and other flight vehicles, with real strength in both our mission and the core GNC disciplines. We design, prototype, test, iterate and learn together. Because GNC is central to safe flight, we drive the concepts of operations and many of the system-level analyses on the program. We also have a lot of freedom. We trust our engineers to choose their own tools and approaches for the problems they face, and to own the results. If you want the autonomy to solve hard problems your way, with a team of strong engineers, you will feel at home here.
US, WA, Seattle
Do you want to make AI provably correct for chip design and verification? We're hiring an Applied Scientist to combine large language models with formal methods such as Lean 4 and turn AI from a helpful assistant into a trusted engineering tool for silicon. Our team builds the formal verification methods behind custom silicon at Amazon scale. As AI agents take on more of chip design and verification, we need people who can make sure their output is checkable, sound and grounded in mathematics. You will research and build systems where LLMs work alongside interactive theorem provers, SMT solvers and model checkers. Engineers write the specifications. AI refines them against RTL, generates proof steps and closes proofs at machine speed. Every result is checked by a formal kernel, not trusted on faith. Work at the frontier of AI reasoning, theorem proving and hardware verification Ship solutions that directly affect real chips running in AWS data centers Publish and help shape how the industry formalizes AI use in hardware engineering Key job responsibilities Design and build LLM-driven systems for formal verification and design tasks, including proof generation, invariant and lemma discovery, spec-to-RTL refinement, and counterexample analysis Develop frameworks in Lean 4 (and related tools such as SMT solvers, model checkers and commercial FV engines) that formally check AI-generated artifacts, so they are correct by construction rather than just plausible Formalize hardware semantics, ISAs and microarchitectural properties in Lean 4 to create machine-checkable foundations for AI agents to reason over Track and evaluate state-of-the-art AI work (LLM agents, autoformalization, neural theorem proving, reinforcement learning from proof feedback) and turn promising research into production tools Build benchmarks and evaluation methods that measure AI soundness, proof success rate and engineering impact on real designs Partner with FV and design engineers to find the highest-leverage problems and deploy solutions into active chip programs Publish research, file patents and represent Amazon in the formal methods and AI communities Mentor engineers and scientists on applying formal AI methods
US, CA, Pasadena
We are seeking Applied Scientists to join Amazon. In this role, you will develop and deploy the science of safe autonomy on highly dynamic robots, advancing the frontier through both foundational research and realization on hardware. You will contribute to achieving safe autonomy across three key areas: control barrier function (CBF) theory for robust and performant safety on hardware; safe reinforcement learning for agile whole-body control; and layered safety filters that interface with learning, perception, and semantic reasoning systems. You will validate your work experimentally on state-of-the-art robotic platforms — removing bottlenecks to deployment and enabling robots to safely operate around humans. You will work with the inventor of CBFs, as well as top scientists and engineers at Amazon developing the next generation of safe autonomy. Key job responsibilities -Advance the science of safe autonomy from formal foundations to the integration with learning and perception to validation on hardware — with particular emphasis on methods that bridge these domains. -Test and validate hardware, with a focus on next generation robotic systems. Including locomotion, manipulation and loco-manipulation. -Leverage the deployments on hardware to validate the underlying science, identifying gaps between theory and practice that drive the next cycle of research. hardware to validate the underlying science, identifying gaps between theory and practice that drive the next cycle of research. -Publish research at top-tier robotics, control, and ML venues, and contribute to Amazon’s scientific reputation in advanced robotics. -Collaborate with SAF Lab and Amazon production teams to move research into robots deployed at Amazon scale. A day in the life Amazon offers a full range of benefits that support you and eligible family members, including domestic partners and their children. Benefits can vary by location, the number of regularly scheduled hours you work, length of employment, and job status such as seasonal or temporary employment. The benefits that generally apply to regular, full-time employees include: -Medical, Dental, and Vision Coverage -Maternity and Parental Leave Options -Paid Time Off (PTO) -401(k) Plan If you are not sure that every qualification on the list above describes you exactly, we’d still love to hear from you! At Amazon, we value people with unique backgrounds, experiences, and skillsets. If you’re passionate about this role and want to make an impact on a global scale, please apply! About the team The Safe Autonomy Frontiers (SAF) Lab is the first industry research lab dedicated to safe autonomy, founded by the inventor of control barrier functions. We are developing a universal safety layer for the next generation of robotic systems: mobile robots, manipulators, quadrupeds, and humanoids. Here you will advance performant safety for highly dynamic robots — CBF theory integrated with perception and learning, evaluated on next-generation platforms — and your work will underpin robots operating alongside people at Amazon’s unprecedented scale.
US, MA, North Reading
Robotics Applied Scientist II Intern/Co-op (2027) Are you passionate about robotics and research? Do you want to solve real customer problems through innovative technology? Do you enjoy working on scalable research and projects in a collaborative team environment? Do you want to see your science solutions directly impact millions of customers worldwide? We’re looking for current PhD students with a passion for robotic research and applications to join us in 2027 as Robotics Applied Scientist II Interns/Co-ops to shape the future of robotics and transform automation at Amazon’s scale. By applying to this role, you will be considered for Robotics Applied Scientist II Intern/Co-op (2027) opportunities across various Robotics teams at Amazon with different robotics research focus. Internship positions are available for multiple locations, durations (3 to 6+ months), and year-round start dates (winter, spring, summer, fall). Join us at the forefront of Applied Science, where your contributions will shape the future of Robotics! Key job responsibilities As a Robotics Applied Scientist II Intern/Co-op, you will work alongside global experts to advance the state of the art in robotics and artificial intelligence, tackling innovative problems that bridge the gap between research and real-world robotics implementation to innovate and build solutions on behalf of our customers. We are particularly interested in students with specialization and expertise in one or more of the following research areas: - Robotics - Artificial Intelligence - Robotics Manipulation (e.g., robot arm, grasping, dexterous manipulation, end of arm tools/end effector) - Autonomous Mobile Robots - Mobile Manipulation - Multi-Robot Coordination - Autonomous Navigation - Locomotion - Motion/Path Planning - Controls - Perception - Sensing - Robot Learning - Machine Learning - Computer Vision - Large Language Models (LLM), Vision-Language-Action Models (VLA), Vision-Language-Models (VLM) - World Models - Human-Robot Interaction - Robotics Simulation - Optimization - And related! Key responsibilities may include: - Develop novel, scalable algorithms and/or modeling techniques that advance the state-of-the-art in science areas such as robotics, artificial intelligence, or related - Tackle challenging, innovative research problems within robotics, artificial intelligence, or related that bridge the gap between theoretical advancement and practical robotics implementation - Collaborate with cross-functional teams to solve complex problems, and to communicate research findings clearly, leveraging your expertise in robotics, artificial intelligence, or related - Demonstrate the ability to work independently, thrive in a fast-paced, ever-changing environment - Leverage AI-powered tools where applicable to accelerate research, experimentation, and prototyping. - Critically review and validate outputs from AI tools and automated systems. About the team Explore Robotics Teams at Amazon At Amazon, robotics is where innovative research meets real-world deployment. There are many different specialized robotics teams, including the few highlighted below. Learn more about Robotics at Amazon here: https://www.amazon.science/research-areas/robotics Fulfillment Technologies and Robotics We build dynamic partnerships between people and intelligent machines. This intricate collaboration helps Amazon fulfill orders with unmatched accuracy. Working in symphony with our robotic technology, you’ll have the opportunity to extend your technical capabilities by working alongside some of the industry’s most advanced technologies. This includes our fleet of autonomous mobile robots, sophisticated control software, and technologies like language perception, machine learning, object recognition, and semantic understanding of commands. These technologies help us deliver an ever-improving customer and employee experience, as well as improve the safety of our facilities. Amazon Robotics Amazon Robotics is at the epicenter of robotics innovation. We design, test, and build mobile robots and mechatronic systems. Our offices, labs, test floors, and factories are close together, which means we can design the hardware, write the code, test, build, and watch it all come to life. Our areas of focus include advancements in autonomous mobility, AI, robotic management software, and much more. Fauna Fauna is focused on building capable, safe, and genuinely delightful robots for everyday life. Fauna’s goal is simple: make robots people actually want to live and interact with everyday. Fauna develops tightly integrated hardware and software systems that make it faster, safer, and more intuitive to create real-world robotic products. Fauna’s work spans the full stack: mechanical design, control systems, dynamic modeling, and intelligent software. The focus is not just functionality, but overall experience. Amazon Fulfillment Technologies Amazon Fulfillment Technologies develops tech solutions that create network optimizations for all Amazon sites worldwide. We own network infrastructure and devices in all buildings and labor planning and scheduling technology for fulfillment centers in North America and Europe.
US, WA, Seattle
MULTIPLE POSITIONS AVAILABLE Employer: AMAZON.COM SERVICES LLC Offered Position: Applied Scientist III Job Location: Seattle, Washington Job Number: AMZ10037670 Position Responsibilities: Participate in the design, development, evaluation, deployment and updating of data-driven models and analytical solutions for machine learning (ML) and/or natural language (NL) applications. Develop and/or apply statistical modeling techniques (e.g. Bayesian models and deep neural networks), optimization methods, and other ML techniques to different applications in business and engineering. Routinely build and deploy ML models on available data, and run and analyze experiments in a production environment. Identify new opportunities for research in order to meet business goals. Research and implement novel ML and statistical approaches to add value to the business. Mentor junior engineers and scientists. Position Requirements: Master’s degree or foreign equivalent degree in Computer Science, Machine Learning, Engineering, or a related field and two years of research or work experience in the job offered, or as a Research Scientist, Research Assistant, Software Engineer, or a related occupation. Employer will accept a Bachelor’s degree or foreign equivalent degree in Computer Science, Machine Learning, Engineering, or a related field and five years of progressive post-baccalaureate research or work experience in the job offered or a related occupation as equivalent to the Master’s degree and two years of research or work experience. Must have one year of research or work experience in the following skill(s): (1) programming in Java, C++, Python, or equivalent programming language; and (2) conducting the analysis and development of various supervised and unsupervised machine learning models for moderately complex projects in business, science, or engineering. Amazon.com is an Equal Opportunity-Affirmative Action Employer – Minority / Female / Disability / Veteran / Gender Identity / Sexual Orientation. 40 hours / week, 8:00am-5:00pm, Salary Range $167,100/year to $226,100/year. Amazon is a total compensation company. Dependent on the position offered, equity, sign-on payments, and other forms of compensation may be provided as part of a total compensation package, in addition to a full range of medical, financial, and/or other benefits. For more information, visit: https://www.aboutamazon.com/workplace/employee-benefits. #0000