- Design, build, and operate the high-throughput, low-latency services that receive bid requests, execute Liftoff’s bidding logic, and respond to ad exchanges in real time.
- Improve the performance, scalability, and reliability of systems that process millions of requests per second under strict latency constraints.
- Develop and optimize GPU-powered inference services that execute neural network models using TensorRT.
- Profile end-to-end inference pipelines to identify and eliminate bottlenecks in GPU utilization, batching, memory access, serialization, networking, and request handling.
- Partner with machine learning engineers to productionize new model architectures and features, translating modeling requirements into efficient and reliable serving implementations.
- Design and operate large feature store fleets that provide low-latency access to real-time and precomputed features.
- Develop benchmarking and performance-analysis tools that help engineers compare models, understand latency and throughput trade-offs, and identify regressions before deployment.
- Improve experimentation tooling that allows machine learning teams to orchestrate offline training runs, evaluate candidate models, and maintain model leaderboards.
- Own changes through their full lifecycle—from system design and implementation to testing, deployment, observability, capacity planning, incident response, and continued optimization.
Machine LearningAlgorithmsData Structures+1 more