About this role
Amazon DynamoDB is a fully managed NoSQL database that serves more than 1 million customers and delivers single-digit millisecond performance at any scale. It supports individual tables over 200TB and sustains over half a million requests per second for hundreds of customers, with up to 99.999% availability. Behind that scale sits a large fleet of capacity that must be placed and balanced continuously. We are looking for an Applied Scientist to advance the science of capacity utilization and data placement across the DynamoDB fleet. You will work backwards from customer experience and fleet economics to find where capacity is used inefficiently, where scaling bottlenecks constrain the service, and where smarter data placement can raise utilization without degrading latency or availability. You will turn these findings into models and algorithms that inform capacity profile decisions and placement policy. You will partner closely with the DynamoDB teams to bring your inputs into production decisions. This is a customer-obsessed science role for a self-driven scientist. Many of the problems are not yet well defined and no textbook solution exists. You will frame the problem, extend state-of-the-art approaches or invent new ones, and drive the work to production impact with a strong bias for action. Key job responsibilities Identify capacity usage optimization opportunities across the DynamoDB fleet. Quantify the customer and cost impact of each opportunity. - Model the scaling bottlenecks of the service and characterize how they constrain placement and utilization. - Develop data placement approaches that balance customer experience (latency, availability, throughput headroom) against optimal capacity utilization. - Partner with the DynamoDB performance team to incorporate your inputs into capacity profile decisions and placement policy. Validate impact with production data. - Build components that integrate directly into production systems or that directly support the large systems making placement and capacity decisions. - Scrutinize the performance of your algorithms and software during implementation. Resolve root causes and leave systems easier to maintain. - Author or co-author papers for internal or external peer-reviewed venues when the work is novel and business considerations allow.