Member of Technical Staff (AI Inference Engineer)

London. Posted 7 months ago

The role

This person will build and run the inference engine behind Perplexity queries. They will deploy model architectures at scale using Rust, Python, CUDA, and CuTe DSL.

What you will do

Support transformer-based retrieval, text-generation, and multimodal models in inference infrastructure
Port in-house CUDA kernels to NVIDIA's CuTe DSL
Develop internal Rust-based inference server
Profile and fix bottlenecks from network ingress through continuous batching and GPU kernels interleaving
Build dashboards, alerts, and automated remediation
Respond to and learn from production incidents

Required qualifications

3+ years of professional software engineering experience with meaningful work on ML inference or high-performance systems
Deep experience with GPU programming and performance work (CUDA, Triton, CUTLASS, or similar)
Understand modern LLM architectures and are able to bring them up reliably in a production environment
Built and operated production distributed systems under real load
Comfortable working across languages and layers: Rust, Python, CUDA/CuteDSL
Familiarity with at least one deep learning framework (PyTorch, JAX, TensorFlow)
Understanding of GPU architectures (memory hierarchy, warp scheduling, tensor cores)
Understanding of common LLM architectures and inference optimization techniques (e.g. quantization, speculative decoding, prefill-decode disaggregation)

Nice to have

Any other deep systems programming experience
ML compilers and framework internals: PyTorch internals, torch.compile, custom operators
Distributed GPU communication: NCCL, NVLink, InfiniBand, RDMA libraries, model/tensor parallelism
Low-precision inference: INT8/FP8/FP4 quantization, mixed-precision serving
Profiling and debugging tools: Nsight Compute/Systems, CUDA-GDB, PTX/SASS analysis
Container orchestration: Kubernetes, GPU scheduling, autoscaling inference workloads

About the company

Perplexity builds and runs the inference engine behind every Perplexity query and deploys dozens of model architectures at scale.