Building AI agents that actually work: Introducing Nova Act as a service

For years, we've heard bold claims about how AI will transform work—creating visions of virtual assistants that magically automate away digital drudgery. Yet organizations experimenting with AI agents have encountered a sobering reality: most agents only work sometimes, failing to deliver the reliability needed for mission-critical workflows.

This challenge has been at the heart of our work since establishing the AGI Lab a year ago. In that time, we've made a lot of progress—first launching Nova Act as a research preview and later introducing an IDE extension to streamline the agent building process.

Which brings us to today: Nova Act is now generally available as a service on AWS for building and managing highly reliable AI agents at scale. Nova Act provides the fastest and easiest path to build and manage fleets of agents that are truly enterprise-ready, with customers achieving 90% reliability on UI-based workflows, like updating a customer record in a CRM system. We’ve also added new features like a no-code playground and human-in-the-loop oversight as well as preview capabilities like tool calling  to accelerate builders throughout their journey. Our endgame: deliver trusted teammates for business-critical work.

Amazon Nova Act | A New AWS Service

Reliable agents at scale

Think of an AI agent that submits perfectly formatted bug reports across multiple repositories, updates data in 87 different tools (yes, the ones that don’t have APIs), and even orders celebratory pizza when your team hits a major milestone with everyone's preferred toppings. That's what Nova Act can help you build: highly reliable agents that excel at automating routine browser-based workflows that previously required human attention.

Nova Act prioritizes what developers and users actually need: reliability . Our user research shows that reliability is the critical factor preventing broader agent adoption in enterprise environments. Nova Act achieves high reliability through a fundamentally different training approach. Powered by a custom Nova 2 Lite model, Nova Act has state-of-the-art browser-use capabilities—performing better than other models on benchmarks like WorkArena L1 and REAL Bench V1 and V2.

Nova Act BOLDED COMPRESSED_corrected120325.png
View details of benchmarking methodology and supporting data here.

Unlike other offerings that bolt together separate components, we train the model, SDK, orchestrator, and browser controllers as a single unified system. By training all components together as one system, Nova Act achieves the reliability that's missing when models and tools are developed separately and combined later. To deliver this tightly integrated performance, we train our custom Nova model using reinforcement learning in synthetic environments called RL gyms that simulate real-world tools like CRM systems, travel sites, and project management tools. Nova Act learns in these replica environments via trial and error, like how a chess-playing AI gets to grandmaster status by playing against itself in a chess simulator ad nauseam. To help visualize this, our team built Nova Act Gym, which provides a sanitized set of demo environments akin to our RL gyms.

Example of Amazon Nova Act's reinforcement learning (RL) gyms:

Example of Amazon Nova Act's reinforcement learning (RL) gyms.
Example of Amazon Nova Act's reinforcement learning (RL) gyms.

Built for developers who ship

Great models aren't enough. Developers need a streamlined path from ideation to production. Nova Act delivers exactly that: a complete solution that dramatically simplifies agent development while enabling you to achieve enterprise-grade reliability.

The journey starts in Nova Act Playground, where you can prototype agents in minutes using natural language descriptions of your workflow. There's no downloading or complex configuration required—just describe what you want to automate and start testing immediately.

Once you've established a proof of concept, you can use the Nova Act extension to rapidly refine your agent in IDEs like VS Code, Cursor, or Kiro using an intuitive mix of natural language and Python, giving you precisely the control you need. The notebook style coding interface and embedded browser testing environment enable immediate iteration— catching issues before they reach production through live debugging. What you build locally is what you'll deploy—no translation layer, no surprises.

When you're ready to scale, just click "deploy" in the extension and it will use familiar AWS services to deploy to AWS. We've added a simple command-line tool to pair with the Nova Act SDK as well as a Nova Act console on AWS to round out our developer toolkit, so you can move from prototype to production in a few commands.

After deploying agents to production, you can use the Nova Act Console in AWS to manage activity and review past workflow runs—keeping you in control without the overhead of constant supervision. The end result? A dramatically streamlined agent builder experience where you can stay focused on business outcomes rather than wrestling with complex agent infrastructure.

From Cars to Space: How customers use Nova Act today

The best validation isn't just benchmarks—it's what happens when real developers use Nova Act to solve real problems. Here's what customers are saying about their experience building agents:

From brittle to battle-tested: How Sola Systems achieved high reliability for customers

Sola, a rapidly growing startup backed by a16z and Conviction, provides an agentic process automation platform for enterprises around the world. By integrating Nova Act, Sola has been able to scale their workflow offering while maintaining high reliability across business-critical workflows in healthcare, logistics, and other major industries. "Our customers were spending hundreds of hours per week on repetitive tasks across dozens of systems. Legacy automations broke every time something updated, and custom integrations were too slow to build. Our vision has always been to put automation in the hands of people who actually understand the work, not just technical experts. With powerful models like Nova Act, that vision is now a reality," said Jessica Wu, Co-Founder and CEO of Sola. "Nova Act has been essential to making that work at enterprise scale. It’s given us the reliability, speed, and cost efficiency to handle business-critical processes running hundreds of thousands of times per month where errors aren't acceptable."

Amazon Nova Act | Sola Systems

Enterprise scale at startup speed: How Hertz eliminated its QA bottleneck

Hertz's rental platform processes millions in daily bookings, making website reliability critical to their business. Using Nova Act, Hertz automated manual test cases for their core rental flows—creating, modifying, and canceling reservations—across both modern and legacy web applications. Through the Amazon Nova Act AWS Console, they can now manage these tests as workflows and monitor results in real time, protecting against revenue leakage from even brief website disruptions. "Quality assurance used to be our biggest bottleneck, but Nova Act helps solve that. What used to take weeks now takes hours, accelerating our shipping velocity by 5x and democratizing quality ownership to everyone in SDLC," said Mu Qiao, Sr Director of Software Engineering, Hertz. "The AI-powered automation catches issues before they impact our customers or revenue, better enabling our QA team to keep pace with development and deliver a higher-quality customer experience." Following this initial success, Hertz is expanding Nova Act's coverage across their entire test suite, including mobile applications and automated exploratory testing for comprehensive global coverage.

Customer Testimonial Hertz GA

One script, hundreds of websites: How 1Password scaled secure authentication

1Password used Nova Act to simplify everyday tasks like logging in to web apps, helping users and organizations securely access the tools they rely on while reducing manual steps and potential risks. Because Nova Act is purpose-built for UI understanding, 1Password was able to get a single prompt working across hundreds of diverse websites. "Nova Act’s ability to understand and adapt to complex sign-in patterns opens new possibilities for how people and systems securely access the web," said Nancy Wang, SVP of Engineering at 1Password. "For 1Password, this means delivering the same trusted, human-like login experience our users rely on, now powered by AI and scaled across more environments. As AI becomes a participant in work, not just a tool, security and identity need to evolve with it. That’s where 1Password comes in: serving as the trusted security layer for human, machine, and AI identities.”

Amazon Nova Act | 1Password

No code, no problem: How Amazon Leo automated UX testing

Amazon Leo is targeting zero critical customer bugs across its website and mobile products ahead of its satellite internet service launch. The core constraint: executing thousands of test cases during development across all customer-facing applications and backend services running on multiple platforms. Maintaining separate automation codebases consumes significant engineering capacity and slows down velocity. Amazon Leo leverages Nova Act's agentic test automation framework to eliminate this constraint. Test scenarios are written in natural language, and then executed automatically across web and mobile, through a hybrid execution model that delivers speed, consistency, and repeatability. Initial test runs execute agentically, capturing all the details, and subsequent runs replay deterministically three times faster than the agentic execution with zero AI costs. When UI changes occur, the system adapts automatically, eliminating manual updates that slow development. "At Amazon Leo, we're working hard to delight our customers when we launch our low-Earth orbit satellite-based internet service to customers across the world. That demands flawless execution across every customer touchpoint," said Raj Kizhakkekalathil, Director of Engineering at Leo Commerce. "Nova Act has transformed our quality assurance approach. What previously took weeks of effort to create a comprehensive set of automated test scenarios now happens in minutes. This fundamentally changes how we build and validate customer experiences from design through delivery."

Amazon Nova Act | Amazon Leo

The path forward: From tasks to teammates

Today's launch is a significant milestone, but it's just the beginning of our journey. Real work is collaborative, multi-agent, and multi-person. To simulate these dynamics, we're investing heavily in capabilities that will transform agents from tools that handle narrow tasks to true collaborators. We're laying the foundation for a future where AI agents are reliable enough to handle business-critical work, experienced enough to adapt to changing conditions, and humble enough to ask for help when they need it.

Test it yourself

Get started today at nova.amazon.com/act.

Research areas
  • Machine learning

Related content

US, WA, Redmond
Amazon LEO is an initiative to increase global broadband access through a constellation of 3,236 satellites in low Earth orbit (LEO). Its mission is to bring fast, affordable broadband to unserved and underserved communities around the world. Amazon Leo will help close the digital divide by delivering fast, affordable broadband to a wide range of customers, including consumers, businesses, government agencies, and other organizations operating in places without reliable connectivity. About this role We're looking for a Data Scientist to turn data from our satellite fleet into engineering decisions. Every satellite in orbit sends back a steady stream of telemetry. Every unit was also tested heavily on the ground before launch, and the ground segment watches the satellite every day while it operates. Your job is to connect these sources. You'll find out which in-orbit behaviors trace back to what we measured in the factory, which events repeat across the fleet, and which patterns point to weaknesses in the hardware design. Your findings will directly shape how we design, build, test, and operate the next generation of satellites. A typical problem starts small: a handful of satellites show a drifting temperature, an occasional link dropout, or a counter that creeps up faster than expected. You'll work out whether it's a real effect or noise, whether it's spreading, and what the affected units have in common. They might share a manufacturing lot, a configuration, an orbital environment, or a signature in their factory test history. From there you'll form a hypothesis about the physical cause and argue it in front of hardware, firmware, and reliability engineers who will push back hard. You'll also have to say plainly when the data rules your own idea out. Telemetry alone rarely settles the question, so your first stop is the ground test data we already have. Every unit in orbit has a detailed record from manufacturing and environmental testing, and the fleet as a whole is backed by an even larger population of units still on the ground. You'll dig through that history for the same behavior: did the affected units show early signs on the ground, did they sit near a test limit, does the effect appear across many units under certain conditions? Often the answer is already in data no one has looked at in this light. When it isn't, you'll design the ground experiments that can provide it. That means deciding which units to test and under what conditions, how many runs you need before the result means anything, and what outcome would confirm or refute the theory. You'll work with the engineers who run those tests, analyze the results, and close the question with a clear answer and a recommended action. Beyond individual investigations, you'll build the fleet-level view: baselines for healthy behavior, early warning indicators that catch emerging issues before they cause outages, and reliability estimates that tell us how the population is aging. You'll decide which recurring analyses should become automated monitoring and work with software engineers to make that happen. You'll be the person the team turns to when someone asks "is this normal?" about the fleet. The ideal candidate is curious about how hardware fails and comfortable with messy data from many sources. They can turn "we're seeing something odd on a few satellites" into a rigorous analysis and a clear recommendation. What you'll bring: - Master's degree or above in a quantitative field such as statistics, mathematics, data science, physics, electrical or aerospace engineering, or computer science, or PhD - 3/5+ years of hands-on experience in an applied or industry setting using Python and SQL to analyze large datasets such as time series, telemetry, sensor, or test data - Experience applying statistical methods (hypothesis testing, regression, time-series analysis, anomaly detection) to noisy, incomplete real-world data - Experience building visualizations and dashboards that communicate findings to engineers and decision makers - Experience framing and solving loosely defined problems, from the first question to a recommendation people act on - Ability to work with hardware and software engineers and turn their questions into analyses you can actually run Nice to have: - Experience analyzing telemetry from satellites, spacecraft, aircraft, vehicles, or other fleets of deployed hardware - Experience with reliability and failure analysis methods such as survival analysis, Weibull modeling, failure-mode analysis, and root-cause investigation - Familiarity with electronics or RF hardware, manufacturing test, environmental stress screening (thermal cycling, burn-in, HASS/HALT), or yield analysis - Experience with pattern-discovery techniques such as clustering, change-point detection, and unsupervised anomaly detection - Experience with large-scale data tools (Spark, Athena/Presto, Redshift) and AWS - Experience with visualization tools such as QuickSight, Tableau, Grafana, or Plotly/Dash - Experience working with software engineers on data pipelines: data requirements, schema design, data quality checks, and automating recurring analyses - Experience using AI-assisted tools to speed up analysis and automate repetitive work - A track record of changing engineering design or test decisions with data 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 As a Data Scientist, you will… - Analyze in-orbit telemetry to measure the health and performance of satellite subsystems, and detect anomalies and slow degradation early - Correlate in-orbit behavior with pre-launch data, including manufacturing, production test, and environmental test results, to find which ground measurements predict on-orbit performance. Feed those insights back into test limits and screening strategy - Correlate in-orbit data with what the ground segment sees (gateways, customer terminals) and with failure analysis of lab and returned units - Connect events across the fleet by build lot, component, configuration, firmware version, orbit, and operating conditions, so we can tell systemic issues apart from one-off events - Find patterns that point to hardware design weaknesses, measure their impact on the fleet, and present the evidence to hardware engineering teams with clear recommendations - Build visualizations and dashboards that support decisions in fleet health reviews, anomaly investigations, and design reviews - Work with software engineers to improve the data pipeline: define data requirements, improve data quality, link ground and in-orbit datasets, and automate recurring analyses - Own ambiguous investigations end to end, from forming the hypothesis to reporting the outcome to senior engineering leaders About the team Phased Array Antenna System Validation is the team that decides whether a satellite's phased array antennas (PAAs) are ready to fly. These are the electronically steered transmit and receive panels that form the link between the satellite and the ground. We're part of Amazon Leo, Amazon's low Earth orbit satellite constellation, and we own the test strategy, the automation, and the data pipeline that every PAA goes through on its way to orbit. That path runs from early component-level screening through full-panel functional testing and extended environmental stress testing across temperature, vacuum, and vibration. Our responsibility doesn't end at launch. We also own the analysis of telemetry from PAAs in orbit, and we use what the constellation tells us to improve how we test on the ground. We are a hardware-facing software and test team, and we sit where several disciplines meet. Our test automation directs firmware provisioning, self-tests, and environmental profiles on test stations at manufacturing sites. Over a single run it exercises thousands of measurement points: high-speed serial links, clocking, digital beamformer firmware, RF chains, power, and thermal behavior. Every result is tied to a specific serial-numbered part. That builds a detailed, part-level history of how each panel and component behaved under stress, from the factory floor to launch. A large part of our work is investigation. When something fails, we work out whether the cause is the hardware, the firmware, the test equipment, or our own automation, and we decide what should happen next. Over time we have built a growing catalog of failure signatures and their root causes, internal tools for analyzing long-duration tests, and AI-assisted agents that help turn raw logs into conclusions. We care about evidence over assumption. A label on a result isn't the result itself, and a search that finds nothing doesn't prove the problem has no history. Theories get tested against the data before they're accepted. We support multiple hardware generations at once, so the architecture, the questions, and the tooling keep changing. You'll work closely with antenna hardware, firmware, manufacturing, and reliability engineers who are glad to dig into a hard problem with you. Because we own both ends, from the first ground test to behavior in orbit, the loop is closed within the team. What we learn from the constellation goes straight back into our screens, thresholds, and test sequences, so that the next panel we validate is more likely to keep working in orbit.
US, NY, New York
We are seeking an Applied Scientist to develop and optimize Visual Inertial Odometry (VIO) and sensor fusion systems for our intelligent robots. In this role, you will design, implement, and deploy state estimation and tracking algorithms that enable robots to understand their position and motion in real time, even in challenging and dynamic environments. You will own the full pipeline from algorithm development through embedded deployment, ensuring that perception systems run efficiently on resource-constrained robotic hardware. You will also leverage modern machine learning approaches to push the boundaries of classical perception methods, combining learned representations with geometric techniques to achieve robust, real-time performance. This is a deeply hands-on role. You will work directly with sensors, hardware, and real-world data, while prototyping, testing, and iterating in physical environments. The ideal candidate has strong foundations in VIO and sensor fusion, practical experience optimizing algorithms for embedded platforms, and familiarity with how modern deep learning is transforming perception. Key job responsibilities - Design and implement Visual Inertial Odometry algorithms for robust real-time state estimation on robotic platforms like Sprout - Develop multi-sensor fusion pipelines integrating cameras, IMUs, and other sensing modalities for accurate pose tracking - Optimize perception and tracking algorithms for deployment on embedded hardware (e.g., ARM, GPU-accelerated edge devices) under strict latency and power constraints - Apply modern ML-based perception techniques (learned features, depth estimation, neural odometry) to complement and improve classical geometric approaches - Build and maintain calibration, evaluation, and benchmarking infrastructure for perception systems - Collaborate with hardware, controls, and navigation teams to integrate perception outputs into the robot’s autonomy stack - Lead technical projects from research prototyping through production deployment
US, WA, Bellevue
Amazon Leo is a constellation of Low Earth Orbit satellites that will provide low-latency, high-speed broadband network connectivity to unserved and underserved communities around the world. We are looking for an Applied Scientist to be a founding scientist on the Engineering and R\&D team within Leo Infrastructure and IP Security. The team defends the manufacturing lines, launch sites, and global ground infrastructure behind the constellation from the most sophisticated threat actors on the planet. These requirements create open scientific problems at the intersection of agentic AI, real-time stream processing, graph-based reasoning, and behavioral analytics. You will build the science behind a neurosymbolic reasoning platform and the models that detect the behavior of sophisticated threat actors. This is an R\&D role with a production mandate, where you define the problem rather than solve a pre-scoped one, and every model, detection, and agent workflow you build becomes the system Leo's security teams use to protect the constellation. #### Export Control Requirement Due to applicable export control laws and regulations, candidates must be a U.S. citizen. Key job responsibilities - Design and implement scalable, production-grade neurosymbolic systems that integrate symbolic reasoning over graph-based knowledge representations with LLM agents to deliver reliable, verifiable security outcomes. - Design and run reinforcement learning and fine-tuning pipelines (GRPO, PPO, DPO) to optimize language models for security reasoning, triage, and detection-authoring tasks. - Build behavioral and statistical models that detect threat actor behavior, and design the evaluation frameworks that measure model performance against that behavior before trusting a model in production. - Design and build multi-agent systems that autonomously triage, enrich, and contain security events, including the constrained reasoning, safety guardrails, and validation mechanisms that make automated decisions trustworthy at scale. - Own the end-to-end science lifecycle, from research and experimentation through production deployment, defining the metrics that measure system performance and real-world security impact. - Advance the state of the art through publications at top-tier venues, patents, or open-source contributions, and shape the scientific agenda and research culture from day one. A day in the life You will move between research and production in the same week: framing an ambiguous security problem as a scientific question, prototyping an approach, and partnering with software engineers to ship it as a capability the platform runs continuously. Security engineers on your team translate threat intelligence into the adversary behaviors that matter; you build the models that detect those behaviors and evaluate model performance against them. You will obsess over the two latencies that define the platforms, the time from event to detection and the time from detection to containment action, and design agents and detections that drive both down. You will backtest candidate detections against retained telemetry, review evaluation results before a model or agent capability graduates to automated execution, and deliver scientific artifacts that ship. About the team Leo Infrastructure and IP Security protects the people, facilities, hardware, and supply chain behind a global satellite constellation. The Engineering and R&D team within this organization builds the platforms and tooling the security pillar teams operate on, moving security operations from manual triage to correlation-based detection, automated response, and agentic AI. The team is composed of applied scientists, software engineers, and security engineers working across physical and digital security domains. #### Inclusive Team Culture In Amazon Security, it's in our nature to learn and be curious. Ongoing DEI events and learning experiences inspire us to continue learning and to embrace our uniqueness. Addressing the toughest security challenges requires that we seek out and celebrate a diversity of ideas, perspectives, and voices. #### Training & Career Growth We're continuously raising our performance bar as we strive to become Earth's Best Employer. That's why you'll find endless knowledge-sharing, training, and other career-advancing resources here to help you develop into a better-rounded professional. #### Work/Life Balance We value work-life harmony. Achieving success at work should never come at the expense of sacrifices at home, which is why flexible work hours and arrangements are part of our culture. When we feel supported in the workplace and at home, there's nothing we can't achieve.
IN, KA, Bengaluru
We are looking for a talented Research Scientist to join our team. In this role, you will design, develop, and deploy machine learning and computer vision models that solve real-world problems at scale in the Amazon grocery domain. You will work closely with engineering, product, and business teams to turn complex technical challenges into production-ready solutions, and own the model development lifecycle from experimentation through deployment. You will bring scientific rigor to every stage — from data analysis and model design to evaluation and iteration. This is a high-impact role where your models will directly improve the shopping experience for millions of customers in Amazon grocery stores. Key job responsibilities Design, train, and evaluate computer vision and machine learning models for complex grocery-domain problems including product identification, shelf perception, and in-store scene understanding — iterating rapidly from prototype to production-quality solutions Conduct rigorous exploratory data analysis to characterize domain-specific challenges (image variability, catalog gaps, label noise) and translate findings into actionable modeling decisions Design and execute offline and online evaluation frameworks — defining metrics that capture both model performance and downstream business impact, and diagnosing failure modes to prioritize improvements Build and improve data pipelines and annotation workflows that feed model training, including active learning strategies to maximize label efficiency Communicate technical results, trade-offs, and recommendations clearly to engineering, product, and business stakeholders — connecting model behavior to customer experience outcomes Stay current with state-of-the-art research in computer vision, multimodal learning, and representation learning — evaluating and adapting promising techniques to team-specific problems Contribute to a culture of scientific rigor through reproducible experimentation, thorough documentation, peer code and design reviews, and raising the quality bar for the team A day in the life As a Research Scientist on the GRAISE team, you'll spend your days analyzing model performance from overnight experiments, and prototyping new approaches using multimodal learning with store video and sensor data. You'll present findings to product and business stakeholders, translating technical results into actionable recommendations. Throughout the day, you'll balance rigorous scientific thinking with practical engineering constraints, knowing your work directly improves the shopping experience for millions of customers in Amazon grocery stores. About the team The GRAISE team (Grocery, Retail & In-Store Experience) within World Wide Grocery Store Tech (WWGST) builds foundational AI and machine learning systems that power Amazon's in-store grocery technologies. We develop domain-specific models that solve uniquely complex challenges in grocery — from smart shopping carts and inventory intelligence to personalization and store operations. Our mission is to create technology which makes grocery shopping more convenient, economical, personalized, and enjoyable for customers while empowering retailers with operational efficiency.
US, CA, Sunnyvale
Robots are about to step out of the movies and into our homes. If you're excited about this robotics revolution, come join Amazon's Fauna Robotics. As an Applied Scientist in this org, you'll push the state-of-the-art techniques in Robotics. You'll drive technical excellence in areas such as perception, manipulation, sim2real transfer, reinforcement learning and multi-task learning, designing novel algorithms that bridge the gap between research and real-world deployment. In this role, you will integrate hands-on technical expertise with scientific leadership, ensuring your team delivers robust solutions for dynamic real-world environments. While staying updated with the latest advancements in robotics research, you will leverage your experience and expertise to guide the team towards the most promising direction. Collaborating with a talented group of software engineers and scientists, you will bring your ideas to fruition. Deploying robots in home environments requires you to anticipate challenges not encountered in structured settings like warehouses. Additionally, consumer-grade sensors and actuators necessitate working around their limitations, requiring out-of-the-box innovative ideas. As such, you are someone who truly enjoys thinking about problems using first principles at a holistic and systems level to address the aforementioned challenges. Key job responsibilities - Lead technical initiatives in areas such as robotics foundation models, reinforcement learning and manipulation - - Design experiments to identify the limitations of current state-of-the-art models and developing new models or techniques that can surpass these models - Design and implement novel deep learning architectures that push the boundaries of what robots can understand and accomplish - Mentor fellow scientists while maintaining strong individual technical contributions - Collaborate with engineering teams to optimize and scale models for real-world applications - Influence technical decisions and implementation strategies within your area of focus - Experienced in communicating complex technical work to a non-technical audience - The ability to work with minimal guidance, be proactive and to handle ambiguity and the challenge of quickly evolving goals About the team Fauna Robotics, an Amazon company, is building capable, safe, and genuinely delightful robots for everyday life. Our goal is simple: make robots people actually want to live and interact with in everyday human spaces. We believe that future won’t arrive until building for robotics becomes far more accessible. Today, too much effort is spent reinventing the fundamentals. We’re changing that by developing tightly integrated hardware and software systems that make it faster, safer, and more intuitive to create real-world robotic products. Our work spans the full stack: mechanical design, control systems, dynamic modeling, and intelligent software. The focus is not just functionality, but experience. We’re building robots that feel responsive, expressive, and genuinely useful. At Fauna, you’ll work at the frontier of this space, helping define how robots move, manipulate, and interact with people in natural environments. It’s an opportunity to solve hard problems across hardware and software with a team focused on making robotics accessible and joyful to build. If you care about making robotics real for everyone and building systems that are as delightful as they are capable, we’re interested in hearing from you.
US, WA, Seattle
Economists at Amazon partner closely with senior management, business stakeholders, scientist and engineers, and economist leadership to solve key business problems ranging from Amazon Web Services, Prime, inventory planning, international retail, third party merchants, search, pricing, labor and employment planning, effective benefits (health, retirement, etc.) and beyond. Amazon Economists build econometric models using our world class data systems and apply approaches from a variety of skillsets – applied macro/time series, applied micro, econometric theory, empirical IO, empirical health, labor, public economics and related fields are all highly valued skillsets at Amazon. You will work in a fast moving environment to solve business problems as a member of either a cross-functional team embedded within a business unit or a central science and economics organization. You will be expected to develop techniques that apply econometrics to large data sets, address quantitative problems, and contribute to the design of automated systems around the company.
US, NY, New York
The Sponsored Products and Brands team at Amazon Ads is re-imagining the advertising landscape through industry leading generative AI technologies, revolutionizing how millions of customers discover products and engage with brands across Amazon.com and beyond. We are at the forefront of re-inventing advertising experiences, bridging human creativity with artificial intelligence to transform every aspect of the advertising lifecycle from ad creation and optimization to performance analysis and customer insights. We are a passionate group of innovators dedicated to developing responsible and intelligent AI technologies that balance the needs of advertisers, enhance the shopping experience, and strengthen the marketplace. If you're energized by solving complex challenges and pushing the boundaries of what's possible with AI, join us in shaping the future of advertising. As an Applied Scientist within the Brand Intelligence team, you will be responsible for improving the quality of underlying agentic solutions and their impact within the ads delivery stack. We are looking for a passionate candidate with technical expertise in agentic systems, continual learning, evaluation, information retrieval, Natural Language Processing (NLP), and Large Language Models (LLM). In addition to having hands-on experience in building ML-based solutions, an ideal candidate should be able to create and articulate a customer-centric science vision, show willingness to continuously learn about new scientific approaches, and enjoy operating in startup-like environment.
US, CA, Sunnyvale
Amazon is looking for an Applied Scientist with a strong deep learning background to help build the training environments that teach large language models to reason, use tools, and complete complex, multi-step tasks. Key job responsibilities You will design, build, and evaluate reinforcement learning environments (RL gyms) for LLM training: realistic, verifiable task environments where models learn through interaction and feedback. This includes defining tasks and reward signals, building the infrastructure to run environments at scale, measuring how well environments transfer into model capability, and developing rigorous methods to verify environment quality and correctness. You will work closely with model training teams to understand where environments are most valuable and iterate based on what the models learn. Your work will directly shape products and services that use generative AI. About the team The Foundational AI (FAI) team has a mission to push the envelope in GenAI with LLMs, in order to provide the best-possible experience for our customers.
CH, Zurich
Are you passionate about pushing the boundaries of what robots can do? Amazon is looking for a creative and driven Applied Scientist to help build intelligent robotic manipulation systems powered by reinforcement learning. In this role, you will research, design, and deploy machine learning solutions that enable robots to interact with the physical world in increasingly capable ways. You will tackle complex, ambiguous problems where textbook answers don't exist, working backwards from real-world needs to invent new approaches. Your work will directly influence how automation systems learn, adapt, and perform at Amazon's scale — and you will see your research come to life on real hardware. Key job responsibilities - Research and develop reinforcement learning methods for robotic manipulation, advancing the state of the art in sim-to-real transfer, contact-rich control, and adaptive grasping strategies. - Design, implement, and evaluate machine learning models and algorithms, leveraging deep learning frameworks such as PyTorch or TensorFlow to train policies that generalize across varied manipulation tasks. - Build and maintain production-quality code that integrates scientific prototypes into larger robotics systems, making thoughtful tradeoffs between complexity, performance, and maintainability. - Collaborate with cross-functional engineering and science teams to define simulation requirements, validate model performance, and iterate on solutions informed by real-world hardware feedback. - Contribute to the team's scientific roadmap by identifying new research directions, mentoring fellow scientists, and publishing results at peer-reviewed venues when appropriate. A day in the life You might start your morning reviewing overnight training runs, analyzing reward curves and failure modes from your latest reinforcement learning experiments. After a quick sync with robotics engineers, you shift to refining a simulation environment to close a gap you spotted between sim and real performance. In the afternoon, you pair with a teammate to design a new evaluation framework, then wrap up by drafting a short write-up of your findings to share with the broader team. About the team Our team sits at the intersection of machine learning research and real-world robotics, focused on teaching robots to manipulate objects with precision and adaptability. We combine reinforcement learning, physics-based simulation, and hands-on hardware validation to deliver solutions that work beyond the lab. We value intellectual curiosity, collaborative problem-solving, and a shared commitment to turning ambitious research ideas into tangible impact. If you are excited about building intelligent systems that learn from experience and operate at scale, this is the team for you.
US, WA, Seattle
The charter of the Entertainment Economics team includes working on high-impact projects that require economics expertise and a cross-business perspective. Our team supports Amazon’s diverse entertainment businesses – especially Prime Video, Amazon Music, Amazon MGM Studios, and Gaming (Luna) – through complex causal measurements (observational and experimental) and structural modeling. Some of our team’s areas of science expertise include marketing science, long-term value measurement, and experimentation. As a central economics team not embedded in any particular business, we strive to provide objective analyses that take a customer-obsessed, pan-Amazon perspective. This role partners with the Prime Science team to evaluate and improve the way Amazon allocates Prime subscription revenue across its benefit teams, including entertainment businesses. Prime is Amazon's premiere membership program, with over 200mm members world-wide, bringing in billions in Prime subscription revenue annually. Prime 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: fast and free shipping, entertainment & digital content, reading, healthcare, and grocery. Each of these Prime benefits is supported by a different organization, and each relies on Prime subscription revenue to make critical investment decisions. Our team partners with Prime Science to ensure that these allocations are credible, actionable, and incentive compatible for every benefit. To support this work, this role will create science models, validate them, and productionize them. In addition to stellar economics expertise, this role requires systems thinking (anticipating how complex interconnected models will respond to changes in data or modeling), strategic thinking (supporting decisions that affect 200 million+ customers and set Amazon’s direction for years to come), and scientifically novel thinking (using AI and the latest science advancements to achieve move). The ideal candidate is expert in causal inference with experience working in large, complex structural modeling environments. They are a creative problem-solver, able to identify new approaches that solve multiple problems at once while minimizing churn. They think big, not limiting themselves to what’s been done before, while tactfully accounting for the costs of major changes. They can add structure to incredibly ambiguous business contexts, making defensible assumptions and delivering measurements that drive impactful business decisions. They have effective communication and the skill to translate data-driven findings into actionable insights for senior leaders. Finally, they are empathetic and constructive. Our healthy team culture is supportive and fast-paced, and we prioritize learning, growth, and helping each other to continuously raise the bar. Impact and Career Growth In today’s entertainment landscape, critical decisions are made with data and economic models. You’ll help leaders ask the right questions, and then deliver data-driven answers, helping to create the future of Amazon. You’ll help define a long-term science vision in this space and translate it into an actionable roadmap. This role combines science leadership, organizational ability, technical strength, product focus, and business understanding – a perfect recipe for career growth as an economist in tech. Key job responsibilities • Design and build econometric models, especially causal models, to measure the value of the business and its many features • Develop science products from concept to prototype to production, incorporating feedback from scientists and business partners • Independently identify and pursue new opportunities to leverage economic insights across entertainment and Prime • Write business and technical documents communicating business context, methods, and results to business leadership and other scientists • Serve as a technical reviewer for our team and related teams, including document and code reviews