Kyo Harada — Visage AI / Data Scientist
On-device skin analytics with finance-grade pipelines—CoreML optimization, RAG systems, and real-time evaluation.
- −75% model size
- 3× inference
- −60% RAM
- <5% Δ accuracy
Core Capabilities
On-device Skin Scoring
CoreML-optimized models for real-time skin analysis with privacy-first design.
Evaluation Dashboard
RAG/LLM audit with reproducible benchmarks and actionable recommendations.
Privacy by Design
On-device processing so data never leaves the user's device.
Latest Posts
Operations-first AI Service Design: Privacy, Review, and Launch Constraints
An MVP Is Not Yet a Business Service Why AI Service Design Has to Start with Operations, Privacy, and Launch Constraints The easiest mistake in an AI service...
GTM-attached Development: Designing PoC and LoI as Product Inputs
From MVP to B2B API Product Why PoC and LoI planning must be connected to product definition before launch A working MVP can create a misleading sense of...
Architecture as Contract: Why API Core Beats Demo Polish
API Contracts Before Model Polish Why B2B AI Services Need Repeatability, Monitoring, and Versioning Before They Need Better Demos An MVP can look convincing while still being structurally...