Jinwoo Je | Portfolio

I dealt with a significant escalation due to a potential shift in how the RSR network might merge into the AMZL network. A partner team pushed this initiative and launched a pilot node conversion without reviewing wider implications. Based on my knowledge, this change disrupts current planning and reporting products. I cascaded this to partner teams to ensure tech compatibility, assessed implications within our team's scope, and developed a solution: updating metric definitions and calculations in the processing layer without modifying the underlying dataset. The source table separates raw data into required classifications in different columns, making it feasible to build a solution by updating only the processing layer. This approach supports both scenarios—all last mile nodes merging into a single network or both networks managed separately.
Things move slower in Q4 in planning and tech organizations. In retail, Q4 is peak season, with focus shifting to execution and high-volume support. Ironically, planning work peaks in Q3, then resources shift to execution. I'm factoring this into initiative and task planning throughout the calendar year. Q4 is ideal for deep work with deliverables within my control and fewer dependencies.
I'm focusing more on project conception and prioritization, particularly financial entitlement calculation and placing initiatives in internal and partner team queues for coming months. We previously used entitlement values as black boxes. After a frustrating experience this year, I've decided to master financial modeling and the prioritization process. I did considerable financial modeling at my previous employer (Doosan) on M&A, strategic investment, deals, and corporate planning projects. I was fortunate to learn from brilliant seniors there.
As a side project, I built Claro, an agentic task management system, over the weekend and cancelled my 10- year Todoist Pro subscription. Claro is a CLI tool that lives in terminal, uses SQLite for portability without database servers, and leverages LLM as the main interface with MCP as backend logic.