MLX conversions of microsoft/Phi-4-multimodal-instruct for native vision-language inference on Apple Silicon (M1-M4). 4-bit, 8-bit, and bf16 variants.
AI & ML interests
Privacy-first AI for high-stakes professionals. We build local-inference tools for industries where cloud AI isn't an option — occupational health & safety, legal, healthcare, and general office productivity. Our stack runs entirely on-device using Apple Silicon (MLX) and AMD AI PC hardware, ensuring sensitive client data never leaves the machine. Focus areas: On-device multimodal AI, MLX model optimization & quantization, regulatory compliance automation (COR/SECOR auditing), intelligent document processing, and privacy-preserving speech-to-text. Canadian AI company based in Calgary, Alberta.
Recent Activity
View all activity