Projects
Solutions I've designed or built across consulting engagements and self-built products.
As Solutions Architect, delivered a 0-to-1 AI agent transformation for a joint-stock bank's mobile banking app, supporting 150+ conversational scenarios with small-model-driven product recommendations.
As Solutions Architect, delivered a customer-segment operations analytics agent across 0-to-1 and 1-to-10 stages. In the 0-to-1 stage, built the segmentation analytics agent and integrated it with the profiling system, supporting intelligent analysis of 10,000+ segments; in the 1-to-10 stage, improved agent performance and accuracy while reducing compute cost.
As Solutions Architect, built a 0-to-1 branch operations analytics report-generation agent with a customized report (formatting and layout) generation tool; designed a metrics system on a data lakehouse to support multi-dimensional operational insights.
As Solutions Architect, designed the bank's knowledge-processing system and engineering pipeline, and built 10+ domain knowledge applications based on Agentic Skills.
As Solutions Architect, cross-department coordination to launch a data-driven operations pilot for low-asset customers, leading the development and acceptance of the first version of data analytics and governance standards.
Built customer discovery models using machine learning (7 major categories, 15 sub-segments), landing wealth management marketing strategies. Marketing conversion rate surged 12x (to 13.41%), generating hundreds of millions in revenue.
Designed 4,000+ analytics tags across 3 business directions, defined use cases for overseas, consumer, and automotive finance. Coordinated with Tencent Cloud to deploy the tagging management system pilot.
Designed and built a retail customer analytics BI system: 1 executive dashboard, 17 tracking reports. Built a BI report auto-generation plugin in-house.
Self-built data analytics scheduling platform connecting multi-source data clusters. Running 45 source-layer tasks and 300+ warehouse tasks stably, with ML models deployed via Docker Swarm.
Full-stack self-built investment event process analysis SaaS. Designed metric event-based conversion rules based on the PM4py engine.