Social
Automated reasoning

Better-performing “25519” elliptic-curve cryptography

Automated reasoning and optimizations specific to CPU microarchitectures improve both performance and assurance of correct implementation.

By Torben Hansen, John Harrison
September 10, 2024
12 min read
Share

Cryptographic algorithms are essential to online security, and at Amazon Web Services (AWS), we implement cryptographic algorithms in our open-source cryptographic library, AWS LibCrypto (AWS-LC), based on code from Google’s BoringSSL project. AWS-LC offers AWS customers implementations of cryptographic algorithms that are secure and optimized for AWS hardware.

Two cryptographic algorithms that have become increasingly popular are x25519 and Ed25519, both based on an elliptic curve known as curve25519. To improve the customer experience when using these algorithms, we recently took a deeper look at their implementations in AWS-LC. Henceforth, we use x/Ed25519 as shorthand for “x25519 and Ed25519”.

Graviton chip.png
Related content
Formal verification makes RSA faster — and faster to deploy
Optimizations for Amazon's Graviton2 chip boost efficiency, and formal verification shortens development time.

In 2023, AWS released multiple assembly-level implementations of x/Ed25519 in AWS-LC. By combining automated reasoning and state-of-the-art optimization techniques, these implementations improved performance over the existing AWS-LC implementations and also increased assurance of their correctness.

In particular, we prove functional correctness using automated reasoning and employ optimizations targeted to specific CPU microarchitectures for the instruction set architectures x86_64 and Arm64. We also do our best to execute the algorithms in constant time, to thwart side-channel attacks that infer secret information from the durations of computations.

In this post, we explore different aspects of our work, including the process for proving correctness via automated reasoning, microarchitecture (μarch) optimization techniques, the special considerations for constant-time code, and the quantification of performance gains.

Elliptic-curve cryptography

Elliptic-curve cryptography is a method for doing public-key cryptography, which uses a pair of keys, one public and one private. One of the best-known public-key cryptographic schemes is RSA, in which the public key is a very large integer, and the corresponding private key is prime factors of the integer. The RSA scheme can be used both to encrypt/decrypt data and also to sign/verify data. (Members of our team recently blogged on Amazon Science about how we used automated reasoning to make the RSA implementation on Amazon’s Graviton2 chips faster and easier to deploy.)

Elliptic curve.png
Example of an elliptic curve.

Elliptic curves offer an alternate way to mathematically relate public and private keys; sometimes, this means we can implement schemes more efficiently. While the mathematical theory of elliptic curves is both broad and deep, the elliptic curves used in cryptography are typically defined by an equation of the form y2 = x3 + ax2 + bx + c, where a, b, and c are constants. You can plot the points that satisfy the equation on a 2-D graph.

An elliptic curve has the property that a line that intersects it at two points intersects it at at most one other point. This property is used to define operations on the curve. For instance, the addition of two points on the curve can be defined not, indeed, as the third point on the curve collinear with the first two but as that third point’s reflection around the axis of symmetry.

Elliptic-curve addition.gif
Addition on an elliptic curve.

Now, if the coordinates of points on the curve are taken modulo some integer, the curve becomes a scatter of points in the plane, but a scatter that still exhibits symmetry, so the addition operation remains well defined. Curve25519 is named after a large prime integer — specifically, 2255 – 19. The set of numbers modulo the curve25519 prime, together with basic arithmetic operations such as multiplication of two numbers modulo the same prime, define the field in which our elliptic-curve operations take place.

Successive execution of elliptic-curve additions is called scalar multiplication, where the scalar is the number of additions. With the elliptic curves used in cryptography, if you know only the result of the scalar multiplication, it is intractable to recover the scalar, if the scalar is sufficiently large. The result of the scalar multiplication becomes the basis of a public key, the original scalar the basis of a private key.

The x25519 and Ed25519 cryptographic algorithms

The x/Ed25519 algorithms have distinct purposes. The x25519 algorithm is a key agreement algorithm, used to securely establish a shared secret between two peers; Ed25519 is a digital-signature algorithm, used to sign and verify data.

The x/Ed25519 algorithms have been adopted in transport layer protocols such as TLS and SSH. In 2023, NIST announced an update to its FIPS185-6 Digital Signature Standard that included the addition of Ed25519. The x25519 algorithm also plays a role in post-quantum safe cryptographic solutions, having been included as the classical algorithm in the TLS 1.3 and SSH hybrid scheme specifications for post-quantum key agreement.

Microarchitecture optimizations

When we write assembly code for a specific CPU architecture, we use its instruction set architecture (ISA). The ISA defines resources such as the available assembly instructions, their semantics, and the CPU registers accessible to the programmer. Importantly, the ISA defines the CPU in abstract terms; it doesn’t specify how the CPU should be realized in hardware.

Secure multiparty computation.gif
Related content
Amazon's Tal Rabin wins Dijkstra Prize in Distributed Computing
Prize honors Amazon senior principal scientist and Penn professor for a protocol that achieves a theoretical limit on information-theoretic secure multiparty computation.

The detailed implementation of the CPU is called the microarchitecture, and every μarch has unique characteristics. For example, while the AWS Graviton 2 CPU and AWS Graviton 3 CPU are both based on the Arm64 ISA, their μarch implementations are different. We hypothesized that if we could take advantage of the μarch differences, we could create x/Ed25519 implementations that were even faster than the existing implementations in AWS-LC. It turns out that this intuition was correct.

Let us look closer at how we took advantage of μarch differences. Different arithmetic operations can be defined on curve25519, and different combinations of those operations are used to construct the x/Ed25519 algorithms. Logically, the necessary arithmetic operations can be considered at three levels:

  1. Field operations: Operations within the field defined by the curve25519 prime 2255 – 19.
  2. Elliptic-curve group operations: Operations that apply to elements of the curve itself, such as the addition of two points, P1 and P2.
  3. Top-level operations: Operations implemented by iterative application of elliptic-curve group operations, such as scalar multiplication.
Levels of operations.png
Examples of operations at different levels. Arrows indicate dependency relationships between levels.

Each level has its own avenues for optimization. We focused our μarch-dependent optimizations on the level-one operations, while for levels two and three our implementations employ known state-of-the-art techniques and are largely the same for different μarchs. Below, we give a summary of the different μarch-dependent choices we made in our implementations of x/Ed25519.

  • For modern x86_64 μarchs, we use the instructions MULX, ADCX, and ADOX, which are variations of the standard assembly instructions MUL (multiply) and ADC (add with carry) found in the instruction set extensions commonly called BMI and ADX. These instructions are special because, when used in combination, they can maintain two carry chains in parallel, which has been observed to boost performance up to 30%. For older x86_64 μarchs that don’t support the instruction set extensions, we use more traditional single-carry chains.
  • For Arm64 μarchs, such as AWS Graviton 3 with improved integer multipliers, we use relatively straightforward schoolbook multiplication, which turns out to give good performance. AWS Graviton 2 has smaller multipliers. For this Arm64 μarch, we use subtractive forms of Karatsuba multiplication, which breaks down multiplications recursively. The reason is that, on these μarchs, 64x64-bit multiplication producing a 128-bit result has substantially lower throughput relative to other operations, making the number size at which Karatsuba optimization becomes worthwhile much smaller.

We also optimized level-one operations that are the same for all μarchs. One example concerns the use of the binary greatest-common-divisor (GCD) algorithm to compute modular inverses. We use the “divstep” form of binary GCD, which lends itself to efficient implementation, but it also complicates the second goal we had: formally proving correctness.

Secure multiparty computation.gif
Related content
Computing on private data
Both secure multiparty computation and differential privacy protect the privacy of data used in computation, but each has advantages in different contexts.

Binary GCD is an iterative algorithm with two arguments, whose initial values are the numbers whose greatest common divisor we seek. The arguments are successively reduced in a well-defined way, until the value of one of them reaches zero. With two n-bit numbers, the standard implementation of the algorithm removes at least one bit total per iteration, so 2n iterations suffice.

With divstep, however, determining the number of iterations needed to get down to the base case seems analytically difficult. The most tractable proof of the bound uses an elaborate inductive argument based on an intricate “stable hull” provably overapproximating the region in two-dimensional space containing the points corresponding to the argument values. Daniel Bernstein, one of the inventors of x25519 and Ed25519, proved the formal correctness of the bound using HOL Light, a proof assistant that one of us (John) created. (For more on HOL Light, see, again, our earlier RSA post.)

Performance results

In this section, we will highlight improvements in performance. For the sake of simplicity, we focus on only three μarchs: AWS Graviton 3, AWS Graviton 2, and Intel Ice Lake. To gather performance data, we used EC2 instances with matching CPU μarchs — c6g.4xlarge, c7g.4xlarge, and c6i.4xlarge, respectively; to measure each algorithm, we used the AWS-LC speed tool.

In the graphs below, all units are operations per second (ops/sec). The “before” columns represent the performance of the existing x/Ed25519 implementations in AWS-LC. The “after” columns represent the performance of the new implementations.

Signing new.png
For the Ed25519 signing operation, the number of operations per second, over the three μarchs, is, on average, 108% higher with the new implementations.
Verification.png
For the Ed25519 verification operation, we increased the number of operations per second, over the three μarchs, by an average of 37%.

We observed the biggest improvement for the x25519 algorithm. Note that an x25519 operation in the graph below includes the two major operations needed for an x25519 key exchange agreement: base-point multiplication and variable-point multiplication.

Ops:sec new.png
With x25519, the new implementation increases the number of operations per second, over the three μarchs, by an average of 113%.

On average, over the AWS Graviton 2, AWS Graviton 3, and Intel Ice Lake microarchitectures, we saw an 86% improvement in performance.

Proving correctness

We develop the core parts of the x/Ed25519 implementations in AWS-LC in s2n-bignum, an AWS-owned library of integer arithmetic routines designed for cryptographic applications. The s2n-bignum library is also where we prove the functional correctness of the implementations using HOL Light. HOL Light is an interactive theorem prover for higher-order logic (hence HOL), and it is designed to have a particularly simple (hence light) “correct by construction” approach to proof. This simplicity offers assurance that anything “proved” has really been proved rigorously and is not the artifact of a prover bug.

Diagram of a circuit with a multiplicative depth of two
Related content
Building machine learning models with encrypted data
New approach to homomorphic encryption speeds up the training of encrypted machine learning models sixfold.

We follow the same principle of simplicity when we write our implementations in assembly. Writing in assembly is more challenging, but it offers a distinct advantage when proving correctness: our proofs become independent of any compiler.

The diagram below shows the process we use to prove x/Ed25519 correct. The process requires two different sets of inputs: first is the algorithm implementation we’re evaluating; second is a proof script that models both the correct mathematical behavior of the algorithm and the behavior of the CPU. The proof is a sequence of functions specific to HOL Light that represent proof strategies and the order in which they should be applied. Writing the proof is not automated and requires developer ingenuity.

From the algorithm implementation and the proof script, HOL Light either determines that the implementation is correct or, if unable to do so, fails. HOL Light views the algorithm implementation as a sequence of machine code bytes. Using the supplied specification of CPU instructions and the developer-written strategies in the proof script, HOL Light reasons about the correctness of the execution.

CI integration.png
CI integration provides assurance that no changes to the algorithm implementation code can be committed to s2n-bignum’s code repository without successfully passing a formal proof of correctness.

This part of the correctness proof is automated, and we even implement it inside s2n-bignum’s continuous-integration (CI) workflow. The workflow covered in the CI is highlighted by the red dotted line in the diagram below. CI integration provides assurance that no changes to the algorithm implementation code can be committed to s2n-bignum’s code repository without successfully passing a formal proof of correctness.

The CPU instruction specification is one of the most critical ingredients in our correctness proofs. For the proofs to be true in practice, the specification must capture the real-world semantics of each instruction. To improve assurance on this point, we apply randomized testing against the instruction specifications on real hardware, “fuzzing out” inaccuracies.

Constant time

We designed our implementations and optimizations with security as priority number one. Cryptographic code must strive to be free of side channels that could allow an unauthorized user to extract private information. For example, if the execution time of cryptographic code depends on secret values, then it might be possible to infer those values from execution times. Similarly, if CPU cache behavior depends on secret values, an unauthorized user who shares the cache could infer those values.

Our implementations of x/Ed25519 are designed with constant time in mind. They perform exactly the same sequence of basic CPU instructions regardless of the input values, and they avoid any CPU instructions that might have data-dependent timing.

Using x/Ed25519 optimizations in applications

AWS uses AWS-LC extensively to power cryptographic operations in a diverse set of AWS service subsystems. You can take advantage of the x/Ed25519 optimizations presented in this blog by using AWS-LC in your application(s). Visit AWS-LC on Github to learn more about how you can integrate AWS-LC into your application.

To allow easier integration for developers, AWS has created bindings from AWS-LC to multiple programming languages. These bindings expose cryptographic functionality from AWS-LC through well-defined APIs, removing the need to reimplement cryptographic algorithms in higher-level programming languages. At present, AWS has open-sourced bindings for Java and Rust — the Amazon Corretto Cryptographic Provider (ACCP) for Java, and AWS-LC for Rust (aws-lc-rs). Furthermore, we have contributed patches allowing CPython to build against AWS-LC and use it for all cryptography in the Python standard library. Below we highlight some of the open-source projects that are already using AWS-LC to meet their cryptographic needs.

Open-source projects.png
Open-source projects using AWS-LC to meet their cryptographic needs.

We are not done yet. We continue our efforts to improve x/Ed25519 performance as well as pursuing optimizations for other cryptographic algorithms supported by s2n-bignum and AWS-LC. Follow the s2n-bignum and AWS-LC repositories for updates.

Research areas
Tags
About the Author
Torben Hansen
Torben Hansen is an applied scientist with AWS Cryptography.
John Harrison
John Harrison is a senior principal applied scientist in Amazon’s Automated Reasoning Group. He is a maintainer of s2n-bignum and the HOL Light theorem prover.

Related content

Work with us

See more jobs See more jobs
Research Scientist, Pricing Science
US, WA, Seattle
Pricing is one of the most consequential decisions Amazon makes, and the science behind it needs to be causally rigorous, not just predictive. The P2 Optimization Science team builds the machine learning (ML) systems that power Amazon's pricing decisions at scale: demand lift models, customer lifetime value (LTV) frameworks, and the experimentation infrastructure that validates whether pricing changes actually work. We're looking for a Research Scientist to own causal inference at the intersection of ML and pricing experimentation. Our team has identified a real gap: the methodological bridge between econometric analysis and production-scale ML pipelines needs a practitioner who lives in both worlds. You'll build conditional average treatment effect (CATE) estimation models, design analysis workflows for pricing experiments, and develop reusable causal ML infrastructure that the broader team — including non-ML scientists — can rely on. If you're a scientist who wants to tackle hard causal identification problems in a high-stakes production environment, and who finds satisfaction in making rigorous methods accessible to a broader team, this role is for you. Key job responsibilities - Design, train, evaluate, and deploy end-to-end causal ML pipelines for pricing use cases, including CATE estimation models and heterogeneous treatment effect frameworks. - Serve as the team's causal ML methodology expert — defining identification strategies, model selection criteria, evaluation standards, and the tradeoffs between econometric and ML approaches to causal estimation. - Contribute causal analysis methodology to pricing A/B test post-analysis and build reusable tooling that economists and other non-ML scientists can adopt without requiring deep ML expertise. - Assess emerging causal inference methods (e.g., synthetic difference-in-differences, generalized random forests, causal representation learning) for applicability to Amazon's pricing context, and author internal methodology proposals for adoption. - Partner with senior economists on identification strategy and causal assumptions, collaborate with software and data engineering partners on production deployment, and align with product managers on experiment design requirements. A day in the life Your work directly shapes the prices customers see on hundreds of millions of Amazon products. In a given week you might investigate an optimization anomaly in simulation and trace it to an unmodeled market dynamic, then design an offline evaluation framework to benchmark competing approaches before committing to online testing. You'll collaborate with senior economists on identification strategy for a pricing experiment, present a science proposal for incorporating a new competitiveness signal into an optimization system, or work with the experimentation platform team on randomization design. You'll also have opportunities to write up novel scientific findings for submission to venues such as KDD, NeurIPS, or the ACM Conference on Economics and Computation. About the team The P2 Optimization Science team sits at the center of Amazon's pricing decisions. We build the ML systems, causal inference pipelines, and experimentation infrastructure that determine how prices are set across Amazon's global catalog. Our team brings together research scientists, economists, and engineers who collaborate closely to turn rigorous methodology into production-quality systems with measurable business impact. We value scientific rigor, clear communication across disciplines, and a culture where every team member's work is designed to be extensible and useful beyond their own project. If you want your science to directly influence how hundreds of millions of customers experience Amazon, this is the team to join.
Applied Scientist, Center for Quantum Computing
US, CA, San Francisco
Amazon’s Center for Quantum Computing (CQC) is building the next generation of quantum computing hardware. The Center for Quantum Computing (CQC) is our longest-horizon bet and operates at the intersection of hardware and physics. The goal is a quantum computer that could solve in hours what today's most robust machines would take thousands of years to compute, unlocking breakthroughs in myriad of areas ranging from advanced materials for batteries to cryptography and fundamentally-secure data sharing protocols. This program involves R&D in device theory, materials science, nano-scale fabrication, superconducting circuit design, cryogenic refrigeration, and quantum algorithms and applications. Currently, the quantum hardware team is focused on improving the underlying chip fabrication of the quantum processor and the scaling of the control system needed to run large-scale quantum circuits. As part of this effort, the team is expanding its chip fabrication capabilities through the construction of a new cleanroom facility dedicated to superconducting quantum chip fabrication, and working with strategic partners to develop scaled versions of the control electronics, I/O wiring, and cryogenic refrigeratation needed for utility-scale quantum computers. We operate with a startup-like pace inside one of the world's most capable technology organizations, combining the urgency of a competitive frontier with the resources and infrastructure of pan Amazon. We are seeking an Applied Scientist III to join our team and drive the design, construction, and operation of quantum processors. You will work alongside world-class physicists and engineers to push the frontier of quantum computing — from laser system design and atom trapping to high-fidelity gate execution and quantum error correction. This is a rare opportunity to be part of shaping a technology that could define the next era of computing. Key job responsibilities - Design, build, and operate quantum computing experiments, including optical trapping arrays, laser cooling and state-preparation systems, and high-fidelity gate-driving optics - Develop and optimize two-qubit entangling gates using interactions, targeting fidelities at or beyond the current state of the art (>99.5%) - Implement and characterize erasure-detection protocols leveraging the metastable-state structure of ¹⁷¹Yb to achieve hardware-native erasure conversion rates >90% - Contribute to quantum error correction experiments, including surface codes and quantum low-density parity-check (qLDPC) codes, and work toward demonstrating error suppression factors (Λ) well beyond current published records - Design and improve atom-array loading, rearrangement, and continuous replenishment techniques to support registers of thousands to tens of thousands of atoms - Develop control software and calibration routines for high-speed experimental sequences, including real-time feedback and adaptive protocols - Analyze experimental data, build physical models to explain observed behavior, and propose and execute improvements to system performance - Collaborate with theory, software, and hardware engineering teams to co-design quantum architectures, error-correcting codes, and decoder systems - Publish results in leading scientific journals and present at top conferences to maintain Amazon's visibility in the quantum research community
Imitation Learning Engineer, RIVR
CH, Zurich
RIVR, an Amazon company, is building Physical AI by deploying autonomous robots for real-world doorstep delivery. Operating daily in diverse urban environments, RIVR's robots continuously learn from and navigate the millions of scenarios encountered during deliveries. By owning the full stack from software to hardware, RIVR is purpose-built for safety, reliability, and the customer from day one. Imitation learning enables our robot to mimic "expert" behaviors, derived from human demonstrations or algorithmic strategies. By utilizing state-of-the-art generative AI and similar methods, our wheeled-legged robot can significantly enhance its autonomy and manipulation skills. In this role, you will enable robots to autonomously generate actions from demonstrations and real-time sensor data. These processes may also incorporate responses to natural language commands, further advancing the robot's skills. We are seeking an expert in imitation learning and generative AI techniques that directly produce robot behaviors, along with a deep knowledge of both supervised and self-supervised learning algorithms. If you are passionate about pushing the boundaries of AI and eager to deliver innovative solutions, we invite you to join us in shaping the future of intelligent robotics. Key job responsibilities Develop imitation learning algorithms, such as diffusion policies, to enable robots to autonomously execute actions based on demonstrations and real-time sensor data. Design, test, and refine your algorithms to meet the demands of complex real-world autonomy and manipulation tasks. Construct a dataset for the imitation learning algorithm using human demonstrations or automated expert algorithms. Streamline the workflow to efficiently expand the imitation learning dataset with new tasks. Collaborate with the reinforcement learning team to innovate methods that leverage both simulated and real-world data. Implement deployment-ready code for the real robot, optimized for the robot’s computational constraints. Build, lead and mentor an exceptional team of software engineers. Provide expert guidance to product managers and executives for strategic decision-making. Create and maintain documentation, guidelines, and best practices to streamline knowledge sharing.
Principal Applied Scientist, Ads Measurement Science
US, NY, New York
The Measurement Science team in the Measurement, Ad Tech, and Data Science (MADS) organization of Amazon Ads serves a centralized role developing solutions for a multitude of performance measurement products. We create solutions which measure the comprehensive impact of their ad spend, including sales impacts both online and offline and across timescales, and provide actionable insights that enable our advertisers to optimize their media portfolios. We leverage a host of scientific technologies to accomplish this mission, including Generative AI, classical ML, Causal Inference, Natural Language Processing, and Computer Vision. This Principal Applied Scientist leads the technical vision and execution for Amazon Advertising's performance measurement. Our measurement products and tools help customers understand the holistic impact of their spend on Amazon Ads and optimize their portfolios. This is a complex, multi-disciplinary scientific space spanning machine learning, causal inference, statistics and experimentation, natural language processing, and generative AI. As a GenAI-first organization, we build foundational and agentic models that power advertiser use cases across Ads, while empowering our Applied Scientists to directly build and ship products. You will guide a team of scientists with expertise across different domains to come together and innovate on some of the hardest problems in advertising, made more challenging by evolving privacy regulations and technological changes. In measuring the comprehensive impact of advertising, key challenges you will tackle include capturing impact: 1) both for identified and increasingly unidentified web traffic in light of an evolving global privacy landscape; 2) for outcomes across retailers, including both within and beyond Amazon; and 3) for both short- and long-term impacts of advertising. Key job responsibilities * Driving the scientific vision of the teams in your organization and you advise and influence its technical leadership on performance measurement models and products. * Identifying, tackling, and proposing innovative solutions to intrinsically hard, previously unsolved problems. You solve problems pragmatically, applying judgment and experience to balance trade-offs between competing interests. * Bringing clarity to complex problems, probing assumptions, illuminating pitfalls, fostering shared understanding, and guiding towards effective solutions. * Serving and being recognized by internal and external peers as a thought leader in modeled measurement in Ads. You will engage routinely with the broader scientific community, e.g. as a reviewer, program committee member, workshop organizer, or chair at peer-reviewed internal/external conferences. * Leading by example, with your papers, analyses, and code setting the standard in your organization for scientific engineering excellence. * Leading science and design reviews, aligning teams across your organization around coherent algorithmic strategies. * Influencing your team’s science and business strategy by driving one or more team roadmaps contributing to the organization’s roadmap and taking responsibility for some organizational goals. You drive multiple new product features from inception to production launch. * Guiding the career development of others, actively mentoring and educating the larger applied science community on trends, technologies, and best practices. You drive the documentation of scientific innovation. You attract bar-raising talent from academia and industry. * Staying abreast of academic and industry trends, knowing the state of the art, and building on top of existing solutions instead of reinventing when possible. About the team We are a team of scientists across Applied, Research, Data Science and Economist disciplines. You will work with colleagues with deep expertise in ML, NLP, CV, Gen AI, and Causal Inference with a diverse range of backgrounds. We partner closely with top-notch engineers, product managers, sales leaders, and other scientists with expertise in the ads industry and on building scalable modeling and software solutions.
Robotics Applied Scientist
US, CA, Sunnyvale
Amazon is seeking exceptional talent to help develop the next generation of advanced robotics systems that will transform automation at Amazon's scale. We're building robotic systems that combine frontier AI, sophisticated control, and advanced mechanical design to create adaptable automation capable of working safely alongside humans in dynamic environments. This is an opportunity to shape the future of robotics and automation at an unprecedented scale, working with world-class teams pushing the boundaries of what's possible in robotic manipulation, locomotion, and human-robot interaction. We are seeking an Applied Scientist to develop tactile- and force-driven manipulation policies for highly dexterous, multi-fingered robotic hands and grippers. In this role you will research and build policy learning algorithms that enable robotic hands to perform grasping and in-hand manipulation in unstructured, real-world environments, using touch and force feedback as first-class signals rather than afterthoughts. Much of this work is grounded in reinforcement learning. To be successful you need to be highly motivated, dive deep, and deliver to the highest standards. You will demonstrate strong working knowledge of modern policy learning methods, with expertise in taking algorithms from simulation onto real hardware. That means being comfortable working against real physical constraints and treating sim-to-real as a core research problem rather than optimizing for benchmark numbers alone. You will work across tactile signal processing, contact-rich manipulation, sim-to-real transfer, and multi-modal sensor fusion to build robust solutions for autonomous grasping, dexterous re-orientation, and fine motor control. You will bring the desire to learn from new challenges, and the problem-solving and communication skills to work within a highly interactive and experienced team. Key job responsibilities - Design and train manipulation policies that leverage tactile and force feedback for dexterous grasping, in-hand manipulation, and contact-rich tasks, using reinforcement learning, imitation learning, or hybrid approaches. - Drive sim-to-real transfer, closing the gap between simulated and physical performance on real robotic hands. - Shape objectives and control strategies with the hardware in mind, accounting for how motors and actuators actually behave: transmission ratios, torque versus power, backdrivability, and thermal limits. - Build simulation-based and on-robot evaluation frameworks, with benchmarks and metrics that make tactile perception and policy performance systematically comparable across iterations. - Own scientific and technical projects within the tactile sensing and manipulation workstreams, driving from research concept through deployment on physical robotic systems — and, at the senior level, set direction across a workstream. - Collaborate with hardware, mechanical design, firmware, and controls teams — and with partner applied science organizations — to translate research advances into deployable robotic capabilities, and to inform sensor and actuator design decisions with policy-level requirements. - Publish at top-tier venues and build collaborations with the external research community. - Contribute to a strong scientific bar on the team through code and design review, and mentor engineers and interns working on real-world manipulation and sensing problems. A day in the life Your morning might start with a standup alongside hardware and controls engineers reviewing overnight sim-to-real training runs on a multi-finger gripper, debugging why a grasp policy that worked in simulation is slipping on a real sensor array. Mid-morning you join the weekly tactile sensing workstream to align on sensor integration milestones and share early results from a contact-state estimator you have been prototyping. After lunch you spend a focused block iterating on a reinforcement learning reward formulation, testing variations in simulation before queuing runs on the cluster. Later you pair with a mechanical engineer to review sensor placement trade-offs on an upcoming gripper revision, then wrap the day by drafting a short experiment write-up and sketching next steps for a conference submission. No two days look exactly the same: one week you may be collecting teleoperated demonstrations on the physical robot to seed an imitation learning pipeline; the next you could be deep in a codebase refactor to support a new tactile modality. Throughout, you balance hands-on research — writing algorithms, running experiments, analyzing data — with cross-team collaboration and mentoring. The common thread is moving from research insight to working capability on real hardware, with a tight feedback loop between simulation and the physical world. About the team Our is developing next-generation manipulation capabilities for highly dexterous multi-fingered grippers and hands. Our work spans tactile sensor development, manipulation policy learning, and the tight integration of sensing hardware with intelligent control software. We are building systems that can feel, adapt, and manipulate with human-level dexterity. You will join a team working at the frontier of tactile sensing and contact-rich manipulation, helping define how robots perceive and interact with objects through touch. Because we build the hands as well as the policies that drive them, you will have a direct line to the sensor, actuator, and mechanism designs your algorithms depend on — and real influence over how they evolve. This is an opportunity to solve hard problems across perception, learning, and physical interaction alongside a team of scientists and engineers building toward real-world deployment.
Senior Applied Scientist, Amazon Leo, Amazon Leo Networking Software
US, WA, Redmond
Amazon Leo is building a low Earth orbit satellite broadband network to deliver fast, affordable connectivity to customers around the world, including those underserved by traditional infrastructure. Operating a live constellation and its ground network at scale means service disruptions are an inevitable reality, and how quickly we detect them, understand them, and prevent them from recurring defines the trust we earn as a connectivity provider. We are seeking a talented Senior Applied Scientist to build the next generation of observability and availability science for the Leo network. Leo already detects outages, notifies affected customers, and restores service. The frontier is prediction and prevention: reasoning over complex network and ground-segment telemetry to autonomously detect anomalies before they reach customers, attribute root cause automatically across a sprawling event surface, and rationalize a large, still-maturing metric set so responders trust what the system tells them. In this role you will work backwards from the needs of customers and operations teams. You will own the scientific direction end to end, partnering closely with network engineers, the Network Operations Center, and the Incident Management team, and the Software and Networking orginization to translate a long-term science vision into concrete research and delivery roadmaps. 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 * Own the observability and availability science end to end, from long-term vision to research roadmap to production delivery, working backwards from the needs of customers, the NOC, and network engineering. * Develop time-series and graph-based models with dynamic baselining that detect network and ground-segment anomalies before they cross the High Severity threshold, enabling proactive investigation rather than reactive response. * Reason over complex, multi-system telemetry to attribute cause automatically during live HSEs and LSEs, so Call Leaders and engineers spend time mitigating, not hunting. * Provide the science behind the network's bucketed availability decomposition (geometric and satellite coverage, ground and network segments, customer terminals, payload, and planning), attributing every movement to a true cause rather than noise, so leadership steers on signal and engineering effort lands where it actually moves availability and throughput. * Deliver the models, analyses, and metrics that feed the daily system readiness and validation review consumed by senior leadership and product and engineering, turning availability and throughput data into a defensible, evidence-based read on progress against GA entry criteria. * Own the recurring, fleet-wide "hot topic" investigations that span ASIC, firmware, payload, network, and planning, separating correlated symptoms from true root cause across thousands of nodes, quantifying how much availability or throughput each is costing, and answering whether it is a launch-blocking risk and whether it is improving. * Deliver alert correlation and deduplication, false-positive reduction, and metric-quality scoring so responders trust what the system tells them, directly supporting a large, still-maturing KPI set. * Tie your science to the unit the org actually measures: reduce detection and attribution time, raise availability against the network's bar, and increase the share of events prevented before customer impact. * Partner day to day with the Network Operations Center, Network Engineering, and the Incident Management (Call Leader) program; sit on live HSE/LSE bridges as the science partner and turn event learnings into model improvements. * Feed COEs and RCAs back into detection and attribution so recurring signatures are caught earlier, and help design and score gamedays that stress-test both the network and your models against real, closed events. * Build and deploy production pipelines on AWS (e.g., SageMaker, Bedrock, Glue) with model validation, monitoring, drift detection, and explainability, because operations teams must be able to trust and act on model output under pressure. * Set data-science standards, mentor scientists, and drive best practices across the Leo networking organization as the science function scales. * Produce explainable output and written narratives that let responders and leaders three or more levels up act on your models during and between events.
Data Scientist, OTS - Data ANCHOR Team
US, TX, Austin
The OTS ANCHOR (Analytics, Insights, and Centralized Hub for OTS Reporting) team is part of the WW IT Operations & RME organization, responsible for delivering data solutions that drive operational excellence across Amazon's global IT infrastructure. We build data pipelines, predictive models, develop analytical intelligence frameworks, and create AI-powered solutions that transform how Amazon manages, triages, and resolves IT operational incidents at scale pointing towards the Letting Data Tell the Story north start vision. We are seeking a Data Scientist to help own the end-to-end data science charter for critical operational workstreams — from forecasting IT service demand and capacity planning to designing intelligent severity triaging models and building the analytical backbone for agentic AI solutions. You will work backwards from complex operational problems to create models and solutions that directly impact the efficiency and reliability of Amazon's IT operations worldwide. Key job responsibilities - Design, develop, and deploy machine learning models for IT operations forecasting, including incident volume prediction, capacity planning, and workforce demand modeling - Build and maintain severity classification and intelligent triaging models that process thousands of daily IT service tickets across global regions - Develop anomaly detection systems to identify emerging patterns in operational data, supporting proactive incident prevention - Apply econometric and statistical methods to measure the impact of operational interventions and inform resource allocation decisions - Partner with engineering teams to productionize ML models via automated pipelines using AWS services (SageMaker, Step Functions, Lambda, Redshift) - Design and conduct experiments to empirically validate model performance and operational impact - Contribute to data quality monitoring frameworks and help maintain standards for operational data integrity - Collaborate with cross-functional teams (ServiceNow platform, Decision Intelligence, Field Operations) to translate analytical insights into actionable recommendations - Communicate findings, model performance, and recommendations to stakeholders and senior leaders through written documents and presentations - Contribute to the team's science roadmap, identifying opportunities to apply ML/AI (including Generative AI) to operational challenges A day in the life Amazon Benefits: 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: 1. Medical, Dental, and Vision Coverage 2. Maternity and Parental Leave Options 3. Paid Time Off (PTO) 4. 401(k) Plan
Data Scientist, WWGS SSG Strategic Initiative Development
CN, 31, Shanghai
As WWGS is committed to building an AI-native, agentic organization, the heart of SSG's evolution is to reinvent how we engage sellers with an AI-first mindset. Delivering that ambition at scale demands breakthrough innovation in our foundational capabilities, at ever-faster speed. To meet these priorities with the pace and focus they require, we are standing up a dedicated team, organized as a POD model - a lean, cross-functional team led by a single owner who holds end-to-end decisions and the business outcome, so we can move at the speed the business demands. Our priority is to build the capability to reach the right sellers at scale, anchored on two pillars: Seller Tagging and Good Contact. Through Seller Tagging, we will turn today's fragmented, siloed seller signals into a unified seller intelligence capability that deepens and broadens how well we truly understand every seller. Through Good Contact, we will resolve the seller identity gaps across our channels and systems, so that every engagement and every system that depends on it, can consistently reach the right seller contact. The Data Scientist will be part of the POD team and support the team to establish structured data governance and model training framework, and trail predict models with real business use cases. Key job responsibilities Design and deliver seller segmentation and propensity models at scale — incorporating GMS, category, growth trajectory, engagement signals, and lifecycle stage. Build contact quality scoring and lifecycle management systems (coverage optimization, dormancy detection, reactivation modeling). Define success metrics, experimentation frameworks (A/B, causal inference), and measurement methodology for seller engagement interventions. Productionize ML models and data products — partner with engineering to deploy seller scores, contact quality indices, and recommendation signals. Explore LLM/GenAI applications: automated insight generation from seller data, contact intent classification, and intelligent report synthesis. Serve as the science representative in bi-weekly NBS theme reviews; present findings and proposals to theme Bar Raisers and leadership. Collaborate with BIE team members to democratize analytical outputs via dashboards and self-serve tools. Contribute to cross-marketplace seller behavior analysis supporting Global Expansion strategy (IN, KR, VN, LATAM). Evaluate, integrate, and iterate on AI systems — assess new AI/ML tools, frameworks, and third-party models for applicability to seller intelligence use cases.
2027 Applied Science Internship - Robotics - United States
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.
Principal Applied Scientist, FAIM (Full Funnel Agentic Intelligence and Models)
US, NY, New York
The FAIM (Full Funnel Agentic Intelligence and Models) builds the agentic AI systems behind next-generation advertising. The team develops multimodal foundation models and LLM-driven agents that power shopper understanding, audience-aware creative generation, and agentic advertising workflows. As a Principal Applied Scientist for the FAIM team, you will invent the models that help advertisers achieve desired outcomes. Today's state of the art models can setup ad campaigns to a specification, but do not understand advertising strategy well enough to help with abstract goals that real advertisers ask about like "how can I make more money." This is a rare charter to build foundational optimization science where little exists today, spanning campaign recommendation, cross-product budget allocation, incrementality measurement, and long-term-sales modeling. You will set technical direction for a growing team, partner with engineering and product to take models from research to production at Amazon scale, and directly move advertiser ROAS and new-to-brand growth. In this role you will - Own the science vision for developing evaluations of the Ads Agent - Partner with scientists from across Ads to develop best practices for evaluating agent performance both online and offline - Raise the science bar for all agent evaluations developed in Ads. - Own establishing a data driven foundation for making decisions about when new features are ready to launch in Ads Agent. Key job responsibilities - Define the long-term scientific vision for agent evaluations in Ads - Translate ambiguous advertiser needs and competing objectives into a concrete science roadmap. - Invent, prototype, and productionize machine learning and optimization solutions for joint budget allocation across sponsored ad products, spanning the shopper journey from awareness to purchase. - Develop rigorous approaches to incrementality measurement and long-term-sales modeling that ground optimization in true advertiser value. - Design and lead large-scale experiments and analyses to validate hypotheses and guide product direction. - Partner closely with engineering and product to define technical contracts, data schemas, and serving systems that carry models into production. - Raise the technical bar across science and engineering through mentorship, design reviews, and hands-on collaboration. - Grow scientific talent and publish impactful research internally and at top-tier venues.
See more jobs
View from space of a connected network around planet Earth representing the Internet of Things.
Get more from Amazon Science
Subscribe to our newsletter

Amazon.com | Conditions of Use | Privacy | © 1996-2026 Amazon.com, Inc. or its affiliates

Social