孙铭浩

Led 0-1 product design and front-end development for XiaoMei MCP client, transforming AI dialogue into an AIPC assistant; achieved **25,000+ internal users within 30 days**, earning high praise across departments.

Directed 0-1 business and interaction design, requirements planning, and front-end development for XiaoMei MCP configuration platform, boosting client user growth by **over 30%** post-launch and improving usability.

Co-developed a Perplexity-like AI deep search tool using DeepSeek-R1 within a **5-day extreme development cycle**, achieving **95% core functionality** and optimizing search response to **800ms**, earning commendation at AI OpenDay.

Upgraded the lead scoring model from rule-based to Deep Learning (DL), significantly improving lead scoring accuracy and effectiveness and boosting sales lead conversion and market response speed.

Designed and implemented automated Duplicate and Discard preprocessing flows for SIMO classification model using NLP, enhancing data processing automation and accuracy, and reducing manual handling errors.

Resolved data flow blockages within the company's data closed-loop products, integrating training requirements and unifying dataset structures, boosting R&D training efficiency by **at least 20%** and accelerating product iteration speed.

Led iterations for various CV annotation tools (active learning, 2D Mask, 3D segmentation, SAM2-based video tracking), achieving **91+% multi-class tagging accuracy for 20,000 images** and increasing 2D bounding box annotation speed by **over 160%**.

Managed iterations for platform business functionalities including performance statistics, issue feedback, and permission page reconstruction, significantly improving project manager efficiency in annotation project billing through full business closed-loop implementation.

Actively participated in sales training and customer Q&A for CV annotation tools, contributing to increased sales order volume and resolving multiple customer queries by collecting real-time feedback.