Teaching speech recognizers new words — without retraining

Using lists of rare or out-of-vocabulary words to bias connectionist temporal classification models enables personalization.

In recent years, automatic speech recognition (ASR) has moved to all-neural models. Connectionist-temporal-classification loss functions are an attractive option for ASR (and specifically end-to-end ASR) because they make predictions without conditioning on previous context, thereby yielding simple models with low inference latency.

Unlike earlier, hybrid ASR models, which used lexicons to match phonemes to word candidates, all-neural models are hard to adapt to rare or unfamiliar words. Biasing connectionist-temporal-classification (CTC) models to new words is particularly difficult because of the lack of context: i.e., the model’s prediction at any given time step is independent of the outputs at the previous time steps, the same prediction scheme that enables decoding with low inference latency.

Related content
Accounting for data heterogeneity across edge devices enables more useful model updates, both locally and globally.

This is a problem for ASR applications in which the operational vocabulary is constantly changing, as when new names — say, “Zelenskyy” — enter the conversation, or when users add new names to their address books. Retraining the ASR model on new datasets featuring new words is a prohibitively time-consuming and computationally intensive way to update large models.

In a paper we presented at this year’s Spoken Language Technologies (SLT) Workshop, we describe a method for enabling a CTC model to correctly transcribe new entity names without the need for retraining. The method includes a variety of techniques for biasing the model toward names on a list. These techniques apply to both the model’s encoder, which converts inputs into vector representations, and its beam search decoder, which evaluates candidate output sequences. The techniques can be applied in combination to maximize the likelihood of accurate transcription.

CTC architecture-high-res.png
The architecture of a connectionist-temporal-classification (CTC) model for automatic speech recognition whose output can be biased toward names on an updatable entity list.

On a dataset with difficult medical terminology like names of diseases and medicines, our method improves the ASR model’s F1 score (which factors in both false negatives and false positives) on these entities from 39% in a model without biasing to 62%. Similarly, on a publicly available Vox Populi benchmark that contains recordings of the European Parliament, our method improves the F1 recognition scores of rare entities (names of cities, people, etc.) from 49% to 80% without any retraining of the base ASR model.

Biasing

Our baseline CTC model is an all-neural network that takes frames of audio (snapshots of the signal spectrum across small durations) as input and converts them into a sequence of probability distributions over subword units — word fragments that can be composed into full words. These probability distributions are represented by a weighted graph of possible subword sequences. To rank candidate word sequences, the model decoder uses beam search combined with an external language model (LM), which encodes the probabilities of sequences of words.

Related content
EMNLP papers examine constrained generation of rewrite candidates and automatic selection of information-rich training data.

Encoder biasing 

To bias the CTC model’s encoder, we use a contextual adapter, a separate module that is trained after we have frozen the weights of the base CTC model. The adapter takes the set of rare words in training examples as inputs and learns a mapping between the words’ subword-unit sequences and their audio representations.

In our base network, we use additional CTC losses to train representations from intermediate layers of the encoder (the 6thand the 12th) to produce subword sequences. This enables the model to use approximations of the outputs in previous time steps to influence prediction of the current frame. Our adapter uses a weighted sum of representations from these intermediate layers as audio representations, thereby countering the conditional-independence assumption of CTC models.

At inference time, we use the contextual adapter to embed a list of rare or out-of-vocabulary (OOV) entity names, and at every time frame of the audio, an attention module tries to match the name embeddings with the audio representation. The attention module can also choose to ignore all of the names by attending to a special <no-bias> token. If the audio does contain some entity from the provided list, the probability of the corresponding sequence of subword units is increased.

attention-plot-high-res.jpg
A plot indicating the weight that an attention mechanism gives to names on a custom entity list when encoding audio.

Decoder biasing

We obtained positive results with the following techniques for decoder biasing. All of these techniques are applied directly at inference time:

Related content
Data augmentation makes examples more realistic, while continual-learning techniques prevent “catastrophic forgetting”.

  1. Adaptive subword boosting in beam search decoding: We dynamically boost the probability of a top-k subword sequence if it begins with a subword that appears on the custom entity list. For example, if “Fremont” is one of the custom words, then if the subword “fre” appears, we boost the probabilities of the subsequent subwords “mo” and “nt”. The boosting score for each subword candidate at time step t is determined dynamically by the difference between its log probability and that of the top-1 hypothesis.
  2. Unigram boosting: We boost the probabilities of words on the list of entity names by adding them to the external LM through an OOV/BOOST class, to keep the LM unmodified during inference.
  3. Phonetic-distance-based rescoring: We take the outputs of the intermediate-layer network — which are phones, or phonetic representations of short speech sounds — and perform forced alignment between them and the output of the CTC model. We compute the cost of this alignment and use it to rescore the n-best lists.
  4. Pronunciation-based lexicon lookup: For rare and OOV words, our phone prediction hypotheses are more accurate than our subword predictions. Therefore, we used forced alignment with the phone predictions of the intermediate-layer network to identify the boundaries between words in the phone sequence. If the sequence of phones corresponding to a word is an exact match for the pronunciation of a word in the lexicon, we replace the word with the lexicon entity.
  5. Grapheme-to-grapheme (G2G) techniques: A grapheme is the smallest meaningful unit of written text. We use a table that maps individual graphemes to their multiple possible pronunciations (i.e., phones) to resolve alternative pronunciations of the words on our list of entity names. The probability of predicting the actual word improves with an increase in the number of these G2G variants.

Joint model

Finally, we present a joint model that combines the encoder- and decoder-biasing techniques described above, and as expected, the techniques are complementary to each other and result in additive gains. Conceptually, the encoder-biasing method aids in generating higher-probability scores for the rare subwords it copies, which helps prevent rare subwords from getting pruned during the beam-search decoding of the subword graph. The rare and OOV words get a further boost from the decoder-biasing techniques, which promote the rare-word candidate paths through the graph to top ranking.

We hope our methodology advances the speech community in the direction of zero-shot personalized ASR for CTC models, which are becoming an increasingly prevalent choice for ASR systems.

Research areas

Related content

US, WA, Seattle
Have you ever wanted to solve a mystery or be part of solving a case? Are you fascinated by detective stories or crime shows on TV? Do you love to catch bad actors, build ML models and solve complex problems. If so, working on the Loss Prevention Tech team as a Sr Applied Scientist is the place for you! We detect theft, fraud and organized crime happening across our global supply chain and operations for millions of items, for hundreds of product lines worth billions of dollars of inventory world-wide. We foster new game-changing ideas, creating ever more intelligent and self-learning systems to maximize the cost savings of Amazon's inventory losses. The primary role of a Sr Applied Scientist within Amazon is to address business challenges through building a compelling case, and using data to influence change across the organization. This individual will be given responsibility on their first day to own those business challenges and the autonomy to think strategically and make data driven decisions. Decisions and tools made in this role will have significant impact to the customer experience, as it will have a major impact on all the fraud investigations happening across Amazon operations. Ideal candidates will be a high potential, strategic and analytic graduate with a PhD in ( Research, Statistics, Engineering, and Supply Chain) ready for challenging opportunities in the core of our world class operations space. Great candidates have a history of building fraud detections, detecting organized crime and the ability to use data and research to make changes. This individual will need to be able to work with a team, but also be comfortable making decisions independently, in what is often times an ambiguous environment. Key job responsibilities - Own KPIs that measure fraud management performance and efficiencies. - Detect and automate theft, fraud MOs - Detect organized crime rings and bad actor clusters - Build data or computer vision based ML models - Perform end to end evaluation of operational defects, system gaps, and scaling challenges (both system and operational). - Contribute to the overall fraud management and product development strategies. - Present key learnings and vision to stakeholders and leadership. - Integrate ML detection models via software applications About the team We believe that building a culture that is welcoming and inclusive is integral to people doing their best work and is essential to what we can achieve as a company. We actively recruit people from diverse backgrounds to build a supportive and inclusive workplace. Our team puts a high value on work-live balance. It isn’t about how many hours you spend at home or at work; it’s about the flow you establish that brings energy to both parts of your life. We believe striking the right balance between your personal and professional life is critical to life-long happiness and fulfillment. We are open to hiring candidates to work out of one of the following locations: Seattle, WA, USA
US, TX, Dallas
Amazon is seeking a highly analytical and skilled Data Scientist to join our OpsTech Infrastructure Engineering (OTIE) team. The vision for our organization is to be the invisible scaffolding to provide Amazon’s network and device infrastructure for Global Operations. We deliver flexible, low-touch, cost-efficient infrastructure products by leveraging data, analytics, and automation to build a highly scalable and accessible network. If you are passionate about working with big data and thrive in a collaborative, innovative environment, we want to hear from you. As a Data Scientist, you will be responsible for data exploration and analyses, as well as AI model development. You will collaborate with data engineers to collect, preprocess, and maintain high-quality datasets. You will dive deep into the available data, identifying trends, patterns, and insights to inform AI initiatives. You will design, develop, and implement AI models, including machine learning and deep learning algorithms, to solve complex business challenges, ensuring that these models are optimized for accuracy, scalability, and real-time performance. You will support the deployment of AI models into production environments, ensuring efficient and reliable operation, and own the model performance monitoring, make improvements, and implement retraining strategies. Strong business and communication skills are essential for collaborating with business owners to develop key business questions and build solutions that provide answers and drive change. Key job responsibilities Thinking Big and generating ideas with the stakeholders. Working with customers and cross-functional stakeholder teams to identify, disambiguate, and define problems. Scoping long-term solutions as a series of smaller, more manageable iterations. Creating data science architectures, and building scalable solutions along with the data engineers. Running simulations, measuring performance, building ML models and designing optimization algorithms. Supporting existing models, while thinking about next generation solutions. Keep up-to-date with the latest AI research, technologies, and industry best practices. Share knowledge and promote AI innovation within the team. We are open to hiring candidates to work out of one of the following locations: Dallas, TX, USA
US, WA, Seattle
The Bad Actor Disincentives (BAD) team is responsible for removing the incentive for Bad Actors while accurately and fairly paying millions of third-party sellers along with disrupting the bad actor flywheel and change the economics of abuse within our store. The team works to ensure that bad actors cannot profit from using our services to abuse customers, selling partners and Amazon. While we obsess over customers, we specialize in obsessing over bad actors to identify their friction points and multiply them exponentially in ways that don’t impact good sellers. Our vision is to ensure Bad Actors do not receive a dollar from selling on Amazon and abusing our policies. If we successfully achieve our vision, then Bad Actors will stop committing misconduct on Amazon. This role requires outstanding technical skills, a deep understanding of machine learning approaches, and a passion for melding ML with great user experience/design. You must have a demonstrated ability for optimizing, developing, launching, and maintaining large-scale production systems. As a key member of the team, you will oversee all aspects of the software lifecycle: design, experimentation, implementation, and testing. You should be willing to dive deep when needed, move rapidly with a bias for action, and get things done. You should have an entrepreneurial spirit, love autonomy, know how to deliver, and long for the opportunity to build pioneering solutions to challenging problems. This role will demand resourcefulness and willingness to learn on both the technical and business side. The challenges we take on can involve a mix of large-scale distributed systems, big data technologies, machine learning science, and require a keen sense of customer obsession and long-term strategic thinking. Key job responsibilities You're a former engineer or scientist who can see the bigger picture. While your career is full of individual wins, it is now more fulfilling when your team is able to build, deliver, and impress. You enjoy leading and mentoring others, and want to work on projects that require innovative and creative thinking alongside deep technical problem solving. You challenge yourself and others to constantly come up with better solutions, and can deliver on a technical roadmap that serves our customers and the business optimally. You communicate clearly, and hold yourself and your team to a high bar. We are open to hiring candidates to work out of one of the following locations: Seattle, WA, USA
US, VA, Arlington
Amazon's Global Hiring Science team ensures we match the right people to the right roles, quickly, fairly, and with an amazing experience. To achieve this, we design, implement, and optimize hiring systems experienced by millions of candidates annually. We work in a data-rich, global environment solving complex problems with deep thought, large-sample research, and advanced quantitative methods to deliver practical solutions that make all aspects of hiring more fair, accurate, efficient, and enjoyable. Key job responsibilities We’re developing a new approach to hiring via a multi-year initiative to evolve how we define jobs and candidate qualifications, how we recommend and promote jobs to candidates, and how we help candidates find the roles in which they will be most successful, satisfied, and engaged. To accomplish this, we’ve created a specialized team of experienced industrial-organizational psychologists, applied scientists, engineers, and UX designers. We're looking for an experienced senior research science manager to lead a team of scientists working on this initiative who is equal parts researcher, consultant, and thought leader, with strong expertise in psychometrics, research methodology, and data analysis. In this role, you will collaborate with cross-functional teams to drive research, development, and implementation of innovative hiring technology, evaluation tools, approaches, and methods. The impact of your work will be global and applicable across all of Amazon’s businesses (e.g., AWS, Retail, Logistics, Kindle, and Business Development) and roles (e.g., hourly, technical, professional). A day in the life What you’ll do: • Manage the development and execution of large-scale, highly-visible global research, validation, and hiring optimization projects. • Solve complex, ambiguous measurement, legal defensibility, and experimental design challenges. • Lead the development and research of new content and approaches to assessment (e.g., high fidelity simulation, interactive item types, constructed response). • Apply the scientific method to answer novel research questions. • Influence executive project sponsors and stakeholders across the company. • Drive effective teamwork, communication, collaboration, and commitment across cross-functional groups with competing priorities. • Oversee complex statistical/quantitative analyses with large datasets. About the team We are a team of scientists, and this is an important part of our professional identities. We take our continuing education as well as our contributions to the continuing education of others seriously. To this end, we regularly look for opportunities to engage in reading groups with our peers, present at internal and external conferences, publish our work, and engage in other professional activities in support of our or others development. Learn more about being a scientist at Amazon: https://www.amazon.science. We embrace differences and are committed to furthering our culture of inclusion. Amazon’s culture of inclusion is reinforced within our 14 Leadership Principles, which remind team members to seek diverse perspectives, learn and be curious, and earn trust. We are open to hiring candidates to work out of one of the following locations: Arlington, VA, USA | Seattle, WA, USA
US, WA, Seattle
Are you interested in working with top talent in Optimization, Operations Research and Supply Chain to help Amazon to efficiently match our Devices with worldwide customers? We have challenging problems and need your innovative solutions to make tremendous financial impacts! The Amazon Devices Science team is looking for a Research Scientist with background in Operations Research, Optimization, Supply Chain and/or Simulation to support science efforts to integrate across inventory management functionalities. Our team is responsible for science models (both deterministic and stochastic) that power world-wide inventory allocation for Amazon Devices business that includes Echo, Kindle, Fire Tablets, Amazon TVs, Amazon Fire TV sticks, Ring, and other smart home devices. We formulate and solve challenging large-scale financially-based optimization problems which ingest demand forecasts and produce optimal procurement, production, distribution, and inventory management plans. In addition, we also work closely with demand forecasting, material procurement, production planning, finance, and logistics teams to co-optimize the inventory management and supply chain for Amazon Devices given operational constraints. Key job responsibilities The successful candidate will be a self-starter, comfortable with ambiguity, with strong attention to detail, and ability to work in a fast-paced and ever-changing environment and a desire to help shape the overall business. Job responsibilities include: - Design and develop advanced mathematical, simulation, and optimization models and apply them to define strategic and tactical needs and drive appropriate business and technical solutions in the areas of inventory management and distribution, network flow, supply chain optimization, and demand planning - Apply mathematical optimization techniques (linear, quadratic, SOCP, robust, stochastic, dynamic, mixed-integer programming, network flows, nonlinear, nonconvex programming) and algorithms to design optimal or near optimal solution methodologies to be used by in-house decision support tools and software - Research, prototype and experiment with these models by using modeling languages such as Python; participate in the production level deployment - Create, enhance, and maintain technical documentation, and present to other Scientists, Product, and Engineering teams - Support project plans from a scientific perspective by managing product features, technical risks, milestones and launch plans - Influence organization's long-term roadmap and resourcing, and onboard new technologies onto the Science team's toolbox We are open to hiring candidates to work out of one of the following locations: Seattle, WA, USA
US, WA, Seattle
Are you passionate about solving unique customer-facing problem in the Amazon scale? Are you excited by developing and productizing machine learning, deep learning algorithms and leverage tons of Amazon data to learn and infer customer shopping patterns? Do you enjoy working with a diversity of engineers, machine learning scientists, product managers and user-experience designers? If so, you have found the right match! Fashion is extremely fast-moving, visual, subjective, and it presents numerous unique problem domains such as product recommendations, product discovery and evaluation. The vision for Amazon Fashion is to make Amazon the number one online shopping destination for Fashion customers by providing large selections, inspiring and accurate recommendations and customer experience. The mission of Fit science team as part of Fashion Tech is to innovate and develop scalable ML solutions to provide personalized fit and size recommendation when Amazon Fashion customers evaluate apparels or shoes online. The team is hiring Applied Scientist who has a solid background in applied Machine Learning and a proven record of solving customer-facing problems via scalable ML solutions, and is motivated to grow professionally as an ML scientist. Key job responsibilities Tackle ambiguous problems in Machine Learning and drive full life-cycle Machine Learning projects. Build machine learning models, perform proof-of-concept, experiment, optimize, and deploy your models into production. Run A/B experiments, gather data, and perform statistical tests. Establish scalable, efficient, automated processes for large-scale data mining, machine-learning model development, model validation and serving. Work closely with software engineers and product managers to assist in productizing your ML models. We are open to hiring candidates to work out of one of the following locations: San Diego, CA, USA | San Francisco, CA, USA | Santa Monica, CA, USA | Seattle, WA, USA | Sunnyvale, CA, USA
US, MA, North Reading
Are you excited about developing generative AI and foundation models to revolutionize automation, robotics and computer vision? Are you looking for opportunities to build and deploy them on real problems at truly vast scale? At Amazon Fulfillment Technologies and Robotics we are on a mission to build high-performance autonomous systems that perceive and act to further improve our world-class customer experience - at Amazon scale. We are looking for scientists, engineers and program managers for a variety of roles. The Research team at Amazon Robotics is seeking a passionate, hands-on Sr. Applied Scientist to help create the world’s first foundation model for a many-robot system. The focus of this position is how to predict the future state of our warehouses that feature a thousand or more mobile robots in constant motion making deliveries around the building. It includes designing, training, and deploying large-scale models using data from hundreds of warehouses under different operating conditions. This work spans from research such as alternative state representations of the many-robot system for training, to experimenting using simulation tools, to running large-scale A/B tests on robots in our facilities. Key job responsibilities * Research vision - Where should we be focusing our efforts * Research delivery - Proving/dis-proving strategies in offline data or in simulation * Production studies - Insights from production data or ad-hoc experimentation * Production implementation - Building key parts of deployed algorithms or models About the team You would join our multi-disciplinary science team that includes scientists with backgrounds in planning and scheduling, grasping and manipulation, machine learning, and operations research. We develop novel planning algorithms and machine learning methods and apply them to real-word robotic warehouses, including: - Planning and coordinating the paths of thousands of robots - Dynamic allocation and scheduling of tasks to thousands of robots - Learning how to adapt system behavior to varying operating conditions - Co-design of robotic logistics processes and the algorithms to optimize them Our team also serves as a hub to foster innovation and support scientists across Amazon Robotics. We also coordinate research engagements with academia, such as the Robotics section of the Amazon Research Awards. We are open to hiring candidates to work out of one of the following locations: North Reading, MA, USA | Westborough, MA, USA
US, WA, Bellevue
Inventory Planning and Control (IPC) is seeking an experienced senior data scientist to join its central science team. Our team owns the core decision models in the space of Buying, Placement, and Capacity Control. Our models decide when, where, and how much we should buy, flow, and hold inventories in our global fulfillment network to meet Amazon’s business goals and to make our customers happy. We do this for hundreds of millions of items and hundreds of product lines worth billions of dollars of world-wide for both our Retail and third-party seller business. Our systems are built entirely in-house, for which we constantly develop new technologies in automated inventory planning, prediction, optimization and simulation. Our systems operate at various scales, from real-time decision system that completes thousands of transactions per seconds, to large scale distributed system that optimizes the inventory decisions over millions of products simultaneously. IPC is also unique in that we are simultaneously developing the science and software of inventory optimization and solving some of the toughest computational/operational challenges in production. Our team members have an opportunity to be on the forefront of supply chain thought leadership by working on some of the most difficult problems in the industry with some of the best product managers, scientists, and software engineers in the industry. Key job responsibilities Candidates will be responsible for developing causal, machine learning and data driven models to enhance the various inventory optimization engines that the team owns. The successful candidate should have solid hands-on experience in applying machine learning or causal inference models. They will also be responsible for conducting data driven analysis to facilitate strategic decisions. They require superior logical thinkers who are able to quickly approach large ambiguous problems and develop a practical plan to tackle. Successful candidates must thrive in fast-paced environments, which encourage collaborative and creative problem solving. They are able to measure and estimate risks, and constructively critique peer research. As a senior scientist, you will also help coach/mentor junior scientists in the team. A day in the life The IPC science team contains a large group of scientists with different technical expertise, who will help and collaborate with you on your projects. In this role, you will also work with our internal customers from the Retail, third-party seller and operations departments worldwide. You will understand their challenges and pain points, and help develop data driven solutions that improve how Amazon manages inventory in our global supply chain. You will work closely with the product managers, engineers and other scientists to turn science proposals into production implementation. About the team We are a team of scientists, product managers and engineers focusing on innovation. We promote experimentation and learn by building. We often tackle the hardest problem in the organization and work cross-functionally. We are at the center of developing inventory solutions to support the rapid growth of Amazon's store business. We are open to hiring candidates to work out of one of the following locations: Bellevue, WA, USA
US, WA, Seattle
Amazon Prime is looking for an ambitious Economist to help create econometric insights for world-wide Prime. Prime is Amazon's premiere membership program, with over 200M members world-wide. This role is at the center of many major company decisions that impact Amazon's customers. These decisions span a variety of industries, each reflecting the diversity of Prime benefits. These range from fast-free e-commerce shipping, digital content (e.g., exclusive streaming video, music, gaming, photos), and grocery offerings. Prime Science creates insights that power these decisions. As an economist in this role, you will create statistical tools that embed causal interpretations. You will utilize massive data, state-of-the-art scientific computing, econometrics (causal, counterfactual/structural, time-series forecasting, experimentation), and machine-learning, to do so. Some of the science you create will be publishable in internal or external scientific journals and conferences. You will work closely with a team of economists, applied scientists, data professionals (business analysts, business intelligence engineers), product managers, and software engineers. You will create insights from descriptive statistics, as well as from novel statistical and econometric models. You will create internal-to-Amazon-facing automated scientific data products to power company decisions. You will write strategic documents explaining how senior company leaders should utilize these insights to create sustainable value for customers. These leaders will often include the senior-most leaders at Amazon. The team is unique in its exposure to company-wide strategies as well as senior leadership. It operates at the cutting-edge of utilizing data, econometrics, artificial intelligence, and machine-learning to form business strategies. A successful candidate will have demonstrated a capacity for building, estimating, and defending statistical models (e.g., causal, counterfactual, time-series, machine-learning) using software such as R, Python, or STATA. They will have a willingness to learn and apply a broad set of statistical and computational techniques to supplement deep-training in one area of econometrics. For example, many applications on the team use structural econometrics, machine-learning, and time-series forecasting. They rely on building scalable production software, which involves a broad set of world-class software-building skills often learned on-the-job. As a consequence, already-obtained knowledge of SQL, machine learning, and large-scale scientific computing using distributed computing infrastructures such as Spark-Scala or PySpark would be a plus. Additionally, this candidate will show a track-record of delivering projects well and on-time, preferably in collaboration with other team members (e.g. co-authors). Candidates must have very strong writing and emotional intelligence skills (for collaborative teamwork, often with colleagues in different functional roles), a growth mindset, and a capacity for dealing with a high-level of ambiguity. Endowed with these traits and on-the-job-growth, the role will provide the opportunity to have a large strategic, world-wide impact on the customer experiences of Prime members. Amazon is committed to a diverse and inclusive workplace. Amazon is an equal opportunity employer and does not discriminate on the basis of race, national origin, gender, gender identity, sexual orientation, protected veteran status, disability, age, or other legally protected status. For individuals with disabilities who would like to request an accommodation, visit https://www.amazon.jobs/en/disability/us We are open to hiring candidates to work out of one of the following locations: Arlington, VA, USA | Chicago, IL, USA | Seattle, WA, USA
US, NY, New York
Amazon Advertising is one of Amazon's fastest growing and most profitable businesses, responsible for defining and delivering a collection of advertising products that drive discovery and sales. As a core product offering within our advertising portfolio, Sponsored Products (SP) helps merchants, retail vendors, and brand owners succeed via native advertising, which grows incremental sales of their products sold through Amazon. We deliver billions of ad impressions and millions of clicks and break fresh ground in product and technical innovations every day! The Books Advertising team owns the worldwide advertising business for books, including advertiser and shopper experiences. They develop long-term vision and drive improvements for category relevance, auction dynamics, and ad serving. Additionally, they drive advertiser engagement, represent advertisers' voice, and provide operational support for our programs. This means the team owns all book-specific experiences for Sponsored Products, Sponsored Brands, Sponsored Display, Lock Screen Advertising, the Ads Console, and the Public API. As an Senior Applied Scientist on this team, you will: - Drive end-to-end Machine Learning projects that have a high degree of ambiguity, scale, complexity. - Perform hands-on analysis and modeling of enormous data sets to develop insights that increase traffic monetization and merchandise sales, without compromising the shopper experience. - Build machine learning models, perform proof-of-concept, experiment, optimize, and deploy your models into production; work closely with software engineers to assist in productionizing your ML models. - Run A/B experiments, gather data, and perform statistical analysis. - Establish scalable, efficient, automated processes for large-scale data analysis, machine-learning model development, model validation and serving. - Research new and innovative machine learning approaches. - Recruit Applied Scientists to the team and provide mentorship. Why you will love this opportunity: Amazon is investing heavily in building a world-class advertising business. This team defines and delivers a collection of advertising products that drive discovery and sales. Our solutions generate billions in revenue and drive long-term growth for Amazon’s Retail and Marketplace businesses. We deliver billions of ad impressions, millions of clicks daily, and break fresh ground to create world-class products. We are a highly motivated, collaborative, and fun-loving team with an entrepreneurial spirit - with a broad mandate to experiment and innovate. Impact and Career Growth: You will invent new experiences and influence customer-facing shopping experiences to help suppliers grow their retail business and the auction dynamics that leverage native advertising; this is your opportunity to work within the fastest-growing businesses across all of Amazon! Define a long-term science vision for our advertising business, driven from our customers' needs, translating that direction into specific plans for research and applied scientists, as well as engineering and product teams. This role combines science leadership, organizational ability, technical strength, product focus, and business understanding. Team video https://youtu.be/zD_6Lzw8raE ** Candidates can be based within proximity of NYC, Seattle, Toronto, Arlington County/Virginia (HQ2), or Santa Monica ** We are open to hiring candidates to work out of one of the following locations: Arlington, VA, USA | New York, NY, USA | Santa Monica, CA, USA