iOS Systems Engineer — On-Device LLM Inference Runtime (Swift/C++, Metal, Core ML)

We're building an iOS inference runtime: LLMs running locally on iPhone, with models pushed, swapped and rolled back remotely from a fleet control plane. We need the engineer who builds it. Prototype-grade feasibility work, not production polish. What you'll build A prototype iOS inference runtime (Swift + C/C++), evaluating Core ML and Metal pathways An on-device agent: registration, chunked/resumable/integrity-verified model download, load/swap/rollback, telemetry A test harness app running scripted benchmark suites — latency, memory, thermal state, energy impact — on real devices via AWS Device Farm Must have Strong iOS systems development: Swift with C/C++ interop, memory management under iOS constraints Hands-on Metal compute or Core ML (ideally opinions about both) On-device ML deployment — you've shipped or seriously prototyped models running on iPhones Familiarity with llama.cpp, MLX, Core ML Tools, GGUF quantisation, or similar Nice to have BGTaskScheduler / background-execution expertise MetricKit and energy/thermal instrumentation Model conversion and quantisation SDK or embeddable-framework experience