Former Noniro Tech Core Business Partner Starts Up Vertical Industrial Agents, Secures Tens of Millions in Seed Funding 36氪首发 | 前非夕科技核心业务合伙人创业，做垂域工业智能体，获数千万元种子轮融资

TL;DR

• Shanghai-based startup ZhuiZhi Engineering secured tens of millions in seed funding to develop vertical industrial agents focused on complex manufacturing processes like grinding and welding.
• The company utilizes a proprietary "Industrial Brain + Process Cerebellum" closed-loop control architecture that dynamically adjusts parameters based on real-time perception, moving beyond fixed-program automation.
• Founder Yuan Lin leverages extensive experience from ABB, Huachanda, and Nonirobotics, while the core team includes academic leaders from Shanghai Jiao Tong University and former Tesla Autopilot scientists.
• Commercial validation is underway with a delivered first order to an A-share listed company and multi-million yuan orders in the aviation manufacturing sector.
• The strategic focus is on mastering underlying material mechanics and process capabilities rather than broad industry coverage, enabling cross-industry migration of skills like carbon fiber processing.

Why It Matters

This development highlights a critical shift in industrial AI from generic automation to specialized, physics-aware agentic systems capable of handling non-standardized, high-complexity tasks where human expertise is currently irreplaceable. For practitioners, it underscores the importance of integrating deep domain knowledge (material science) with AI algorithms to create robust, self-adapting industrial solutions that can generalize across different manufacturing contexts.

Technical Details

• Architecture: Implements a self-developed "Industrial Brain + Process Cerebellum" system featuring real-time closed-loop control, autonomous strategy prior judgment, continuous state perception, and dynamic adjustment of process and motion control parameters.
• Core Differentiator: Unlike traditional robots relying on pre-set rules, the system understands material processing mechanisms (e.g., differences between glass, carbon fiber, aluminum, and steel) to define high-quality data collection and self-evolving evaluation systems.
• Hardware Integration: Combines optimized robot bodies, end-effectors, motors, and sensors with the AI control system, ensuring that algorithmic capabilities are grounded in physical hardware constraints and optimizations.
• Product Line: Launched the FAST surface treatment industrial agent, which has been validated in high-precision scenarios such as aviation component manufacturing.
• Business Model: Operates on a three-layer sales structure comprising "Hardware Body + Industrial Brain + Process Data Services."

Industry Insight

• Shift from General to Specific: The success of vertical industrial agents suggests that the next wave of industrial AI value will come from solving specific, high-pain-point processes (like grinding/welding) rather than building general-purpose humanoid robots for all factory tasks.
• Data Quality via Domain Knowledge: True AI generalization in manufacturing depends less on raw data volume and more on the integration of material science and process engineering knowledge to curate high-quality, mechanistic data for training models.
• Cross-Industry Migration Strategy: Companies should prioritize deepening capabilities in specific materials or processes (e.g., carbon fiber processing) over expanding into numerous industries, as these deep competencies can be migrated across diverse sectors like automotive, consumer electronics, and aerospace.