Sectors

About the Role
Own the on-device AI pipeline: fact extraction, entity recognition, Whisper transcription, and local LLM inference. You'll work directly with MLX, Core ML, and Apple's Neural Engine to build fast, private AI that runs entirely on user hardware.

 

Responsibilities

• Architect and optimize MLX inference pipelines for Phi-3 and Llama models.
• Implement and tune Whisper transcription with Core ML acceleration.
• Build fact extraction and entity linking systems that run in <500ms. Design hybrid retrieval (BM25 + vector + re-ranking) for semantic search.
• Optimize model quantization and memory management for consumer hardware.
• Instrument and debug ML pipelines using Sentry and custom telemetry.

Requirements

• 5+ years ML engineering experience shipping production systems.
• Deep experience with on-device ML (Core ML, MLX, ONNX, or similar).
• Strong Python and Swift skills. Hands-on experience with transformer architectures and LLM inference.
• Track record building RAG pipelines and semantic search. BS/MS/PhD in CS, ML, or related field from a top-tier program.
• Prior employment at a recognized tech company or well-funded startup.

Preferred

• Experience with Whisper, speech-to-text systems.
• Published research or open-source contributions in ML.
• Experience with Apple Silicon optimization.

Tech Stack
MLX, Core ML, Swift, Python, Whisper, Phi-3, Llama, SQLite, AWS.