AI Product Manager & AI-native Builder focused on Consumer Agents, Multimodal Experiences, and Model Evaluation.

I design and build AI products end-to-end — from model behavior and evaluation to the interface someone actually lives with — and I write up what I learn as portfolio case studies.

• 🐾 Muse Gadgets Progress Companion — a desktop-pet companion on Meta's open-source Muse Gadgets SDK that shows my real work progress (face, caption, progress ring) via a small file-based state bridge I designed. Full case study, with an interactive prototype: read it here. The bridge source is open: see muse-gadget-state-bridge below.
• 🧪 Agent evaluation work — building evaluation sets and rubrics from real product behavior to measure how consumer agents actually perform, not how demos say they perform.
• 🎲 Board-game agent systems — using board-game systems to probe how agents track state and reason over turns, explored through Unseen and an agent matrix I built for a board game store (recommendations and rules explanation for in-store customers).
• ♟️ Symbolic reasoning explorations — a skill suite for structured/symbolic reasoning tasks.
• 🌙 Unseen — a consumer agent product I'm currently building and iterating on.

Film × Statistics at UCLA — I evaluate models like a statistician and judge experiences like a filmmaker. 3+ years across product, data, growth, and research, including cruise OTA product work and running products as an owner-operator.

• LinkedIn: linkedin.com/in/najiapan