AI Research Scientist

MacPaw is seeking an AI Research Scientist to join their AI&Research Unit, focusing on on-device LLM efficiency and deep architectural explorations for the macOS ecosystem. The role involves fundamental research, collaboration with academic labs, and transitioning prototypes into production.

• Prepare fundamental research proposals within our specialized LLM efficiency and optimization streams.
• Investigate new directions in LLM optimization by surveying relevant academic publications, formulating hypotheses, and running deep-dive experiments.
• Collaborate closely with internal research scientists and external academic labs on joint research projects and scientific publications.
• Work alongside the applied research stream to surface, validate, and transition relevant research prototypes into downstream production environments.
• Contribute to a strong publication record at top-tier international AI conferences.
• Take full ownership of your research niche, driving initiatives from ideation to implementation with a high degree of independence and autonomy.
• Take part in internal knowledge-sharing sessions and weekly paper clubs to consistently improve domain expertise.

• Deep experience in Natural Language Processing (NLP) or a similar machine learning domain, gained through solid academic work, industry experience, or both.
• Strong theoretical and practical understanding of recent LLM optimization techniques (ex. quantization, KV-cache compression, speculative decoding, and distillation).
• Direct hands-on experience with parameter-efficient fine-tuning (including LoRA and other adapters) and mixture-of-experts (MoE) architectures.
• Advanced prototyping skills and a proven ability to implement complex algorithms and architectures directly from academic papers.
• Fluent programming capabilities in Python and modern frameworks such as PyTorch, JAX, or TensorFlow.
• Robust fundamental knowledge of linear algebra, probability theory, and mathematical statistics.
• Upper-intermediate level of English or higher for active scientific writing, international lab collaboration, and conference presentations.
• A track record of publishing original research at international research conferences in AI/ML/SE/HCI domains.
• Practical experience in performance engineering, including code profiling and low-level optimization (GPU, Metal, C++...).
• Hands-on experience training, running, or deploying localized LLM models on device frameworks like MLX.