Explore Agent engineering across products and deployments. Projects share core capabilities including models, harnesses, memory, audio, tools and execution receipts.
Each case explains the task, deployment context and delivered work. Projects without public disclosure approval are described anonymously.
A real-device task entry point with model routing, voice and memory—extending AI beyond one chat into persistent context and executable tools.
Agent Runtime → Tools / Device → Verified Result
Layered evidence, episodes, long-term memory and reusable Agent assets, with retrieval, revision, rollback and controlled access.
Agents collaborate on meetings, research tracking, skills and knowledge organization using isolated workspaces and controlled sharing.
Connect files, web pages, company materials and investment research to specialist agents for reviewable structured reports.
Voice demo and underlying technical delivery for a compact articulated robot collectible project for an international IP brand.
Combines streaming ASR, real-time context and meeting prompts to move from post-meeting summaries to in-meeting support.
From learning diagnostics and AI tutoring to task-first course creation, skill invocation, QA and publishing.
Connect IP knowledge, voice, memory and scene triggers to device interfaces for on-site interaction and staff workflows.
Project experience feeds back into reusable model routing, memory, tools and device interfaces.
User Goals & Business Constraints
Sources, roles and permissions
Planning, models and collaboration
Skill、MCP、Device SDK
Receipts, traces and quality checks
Memory and reusable modules
Whether it is a standalone app, consumer device or business workflow, we can define a minimum verifiable integration together.