We are building AI that runs directly on users’ phones - not in the cloud.
Instead of relying on massive GPU clusters, we take large language models (like GPT or LLaMA) and make them work locally on iOS and Android devices: offline, fast, and without overheating the phone.

This is a very different problem from typical ML roles.
If most ML engineers scale models in the cloud - we compress, optimize, and run them on mobile hardware.

We are looking specifically for engineers who have worked with AI on mobile devices.

What you’ll do
• Take large LLMs and adapt them for on-device inference
• Apply aggressive quantization (INT4, INT8) and model compression
• Run and optimize models on real devices (iPhone / Android)
• Measure latency, thermal throttling, and battery usage
• Convert models for mobile runtimes (CoreML, TFLite, ExecuTorch, llama.cpp)
• Integrate inference into mobile applications

What we’re looking for
• Hands-on experience with on-device / mobile / edge AI
• Experience running models on real mobile devices (not just servers)
• Strong knowledge of model quantization (INT4, INT8, GGUF)
• Experience with:
— ExecuTorch / llama.cpp / MLC-LLM
— CoreML / Apple Neural Engine
— TensorFlow Lite / MediaPipe
• Understanding of mobile chipsets: Snapdragon, ANE, NPUs

Nice to have
• Qualcomm QNN / Hexagon DSP
• ONNX Runtime (mobile)
• Knowledge of KV cache / Flash Attention in mobile context

Important
This role is not about:
• recommendation systems / ranking
• backend ML (RAG, search, chatbots in the cloud)
• GPU/CUDA-heavy server-side ML

For direct applicants only