About the Role
We are looking for an Inference Engineering Manager to lead our AI Inference team. This is a unique opportunity to build and scale the infrastructure that powers Perplexity's products and APIs, serving millions of users with state-of-the-art AI capabilities.
You will own the technical direction and execution of our inference systems while building and leading a world-class team of inference engineers. Our current stack includes Python, PyTorch, Rust, C++, and Kubernetes. You will help architect and scale the large-scale deployment of machine learning models behind Perplexity's Comet, Sonar, Search, Deep Research products.
Why Perplexity?
- Build SOTA systems that are the fastest in the industry with cutting-edge technology
- High-impact work on a smaller team with significant ownership and autonomy
- Opportunity to build 0-to-1 infrastructure from scratch rather than maintaining legacy systems
- Work on the full spectrum: reducing cost, scaling traffic, and pushing the boundaries of inference
- Direct influence on technical roadmap and team culture at a rapidly growing company
Responsibilities
- Lead and grow a high-performing team of AI inference engineers
- Develop APIs for AI inference used by both internal and external customers
- Architect and scale our inference infrastructure for reliability and efficiency
- Benchmark and eliminate bottlenecks throughout our inference stack
- Drive large sparse/MoE model inference at rack scale, including sharding strategies for massive models
- Push the frontier with building inference systems to support sparse attention, disaggregated pre-fill/decoding serving, etc.
- Improve the reliability and observability of our systems and lead incident response
- Own technical decisions around batching, throughput, latency, and GPU utilization
- Partner with ML research teams on model optimization and deployment
- Recruit, mentor, and develop engineering talent
- Establish team processes, engineering standards, and operational excellence
Qualifications
- 5+ years of engineering experience with 2+ years in a technical leadership or management role
- Deep experience with ML systems and inference frameworks (PyTorch, TensorFlow, ONNX, TensorRT, vLLM)
- Strong understanding of LLM architecture: Multi-Head Attention, Multi/Grouped-Query Attention, and common layers
- Experience with inference optimizations: batching, quantization, kernel fusion, FlashAttention
- Familiarity with GPU characteristics, roofline models, and performance analysis
- Experience deploying reliable, distributed, real-time systems at scale
- Track record of building and leading high-performing engineering teams
- Experience with parallelism strategies: tensor parallelism, pipeline parallelism, expert parallelism
- Strong technical communication and cross-functional collaboration skills
Nice to Have
- Experience with CUDA, Triton, or custom kernel development
- Background in training infrastructure and RL workloads
- Experience with Kubernetes and container orchestration at scale
- Published work or contributions to inference optimization research