Demand for large language model (LLM) inference is growing fast, and the engineers who make it faster and more efficient are shaping the future of AI at Amazon. We are looking for a Software Development Engineer to focus on inference performance optimization, kernel tuning, and LLM serving efficiency. You will work across the full software development lifecycle to design, build, and measure systems that reduce latency and improve throughput for production inference workloads. This is a hands-on engineering role where your work directly translates into better performance for internal teams and, ultimately, for Amazon's customers. If you are an engineer who enjoys profiling bottlenecks, writing high-performance code, and finding elegant solutions to hard systems problems, we want to talk to you. Key job responsibilities - Design and implement kernel-level optimizations and inference pipeline improvements that reduce latency and increase throughput for LLM serving workloads. - Profile, benchmark, and tune model inference performance across hardware and software stacks, identifying bottlenecks and delivering measurable gains. - Write well-organized, maintainable code and participate actively in code reviews, providing constructive feedback to peers and mentoring newer engineers on the team. - Collaborate with applied scientists, ML engineers, and infrastructure teams to translate model requirements into production-ready inference solutions. - Evaluate emerging tools, frameworks, and generative AI techniques for inference optimization, recommending adoption where they improve efficiency or quality. A day in the life You might start your morning reviewing benchmark results from an overnight profiling run, then dig into a kernel implementation to shave milliseconds off a critical inference path. After lunch you could pair with an applied scientist to understand a new model architecture's compute profile and plan your optimization approach. Throughout the day you will participate in code reviews, document your design decisions, and share findings with the broader team so that improvements compound across projects. About the team Our team owns SageMaker AI inference at Amazon, and inference performance optimization is becoming increasingly important as LLM workloads scale. Performance optimization is one of our key differentiators, and every improvement we ship has a direct impact on cost, speed, and the experience our customers receive. We value collaboration, technical depth, and a culture where every engineer's perspective is heard. If you want to solve hard systems problems in an environment that rewards curiosity and rigorous engineering, this is the team for you.