Connecting domain knowledge, AI capabilities and real use.
I'm Shihe Zhang, an AI product manager with an interdisciplinary background in psychology and AI-driven innovation. I focus on products for domains where specialist knowledge matters.
I define how AI should work: the features and scoring rules in an assessment, the behaviour and evaluation of a simulated client, or the knowledge and decisions of a conversational Agent. I turn these requirements into workflows, prototypes and checks that a team can implement.
Selected work
Five projects, five product perspectivesAI Projective Assessment Platform ↗
From drawing analysis to an institutional workflow.
Organise specialist criteria into 160+ visual feature dimensions, connecting recognition, scoring, result levels and reporting across individual, batch and customised delivery.
AI Standardized Patient Training & Assessment ↗
Simulated clients, with concrete training feedback.
Turn cases into configurable roles, multi-turn behaviour, two training modes and reference feedback, refined through teacher trials.
Deep Talk Agent ↗
From a vague expectation to confirmable understanding.
Design the Agent architecture and relationship knowledge, linking understanding, dialogue decisions and profiles for exploration and later matching.
EmoCare ↗
A companionship space to explore, stay and return.
Connect emotional expression to daily action through home, district, characters and activities; define architecture, interaction prototypes and MVP requirements.
EEG × Fragrance Experience ↗
Turn a cross-modal idea into a testable experience.
Connect EEG reference signals, composition and subjective preferences through experiment design, service coordination, system proposals and prototypes.
How I work
Define the task before assigning work to AI
I look at where users encounter difficulty, what AI can take on, and which decisions need a user or specialist. Product goals need to become concrete workflows before they can guide features and acceptance criteria.
Turn specialist judgment into knowledge, rules and behaviour
Knowledge, character definitions, prompts and business rules must support the same task. I distinguish what the system needs to know, how it should decide, what action it should take, and how people can inspect its output.
Make the design usable and checkable
I align teams through user flows, prototypes, requirements and rule specifications, then refine the product through testing, professional feedback and institutional requirements.
Shihe Zhang · AI Product Manager
Psychology helps me understand people; product work turns that understanding into systems.
Psychology trained me in research methods, behavioural analysis and individual differences. Studying AI-driven innovation adds a technical, systems and product-strategy perspective. Together, they shape how I translate specialist knowledge into executable AI tasks and how users and professionals can understand, check and correct the output.
My projects span AI-assisted drawing assessment, simulated counselling training, relationship-exploration Agents, digital companionship and an EEG–fragrance proof of concept. I work across requirements and delivery while also designing knowledge structures, scoring rules, prompts, conversational behaviour, prototypes and validation methods.
What I focus on
- AI product mechanisms: translate domain knowledge into inputs, decision conditions, outputs and exception-handling requirements.
- Experience and workflow: consider how end users, professionals and administrators use the same product.
- Validation and iteration: use test outputs, feedback and inspectable cases to guide changes, distinguishing what has been checked from what still needs validation.
Methods and tools
Requirements analysis, user research, prototyping, PRDs, AI rules and evaluation design, experimental design and cross-functional collaboration. I use Figma for product and interaction design; SPSS, R and Mplus for data analysis; and AI coding and prototyping tools to explore implementations.