Inference & Performance Engineer

Designing and implementing high-performance distributed inference systems serving large language models at scale

Profiling and optimizing inference workloads across the full stack from model architecture to GPU kernels

Designing and building production inference serving systems and frameworks

Architecting and managing GPU cluster infrastructure for training and inference workloads

Deploying machine learning models to production and managing their lifecycle from training to serving

Implementing model compression techniques such as quantization, distillation, and knowledge distillation for performance gains

Identifying and eliminating bottlenecks through systematic analysis and root cause investigation

Building and maintaining comprehensive benchmarking suites to measure performance and correctness

Writing and optimizing low-level GPU kernels using parallel computing frameworks

Optimizing resource utilization and cost efficiency across heterogeneous hardware environments

Debugging distributed systems issues across compute, networking, storage, and scheduling layers

Designing request routing, load balancing, and traffic management systems for distributed inference fleets

Building platform abstractions and APIs that hide operational complexity from end users

Building infrastructure-as-code systems that treat hardware provisioning and management as declarative state machines

Designing and implementing continuous performance monitoring and alerting systems

Implementing kernel-level optimizations and performance tuning within operating systems

Collaborating with AI researchers and model teams to co-design architectures optimized for production constraints

Building observability and monitoring systems for production AI infrastructure

Evaluating emerging accelerator hardware and novel AI-specific processor architectures for model workloads

Automating performance regression detection and root cause analysis across distributed training and serving jobs

Optimizing transformer model architectures for edge devices and constrained hardware with strict latency and power budgets

Managing on-call rotations and incident response for mission-critical AI infrastructure

Communicating technical findings clearly and translating performance improvements into business outcomes

Taking ownership of systems from design through production operation and maintenance

Working cross-functionally across research, product, customer engineering, and infrastructure teams

Prioritizing work in fast-moving environments with competing demands and high technical complexity