Make the hierarchy the common language.
I owned the recursive model specification: Organization → Function → Subsystem → Project. A common structure let people roll up and drill down across 130+ programs without rebuilding the logic for each view.
~$4MAnnual infrastructure savings
~90%Forecast accuracy · two-year horizon
130+Chip programs
4 → 2Weeks per planning cycle
AN EXPLANATORY VIEW
See demand before it becomes a constraint.
Forecasting turned infrastructure planning from a spreadsheet exercise into a scenario the team could discuss.
SCHEMATIC · NOT PRODUCTION DATA
Start with the
actual problem.
Capacity planning was a four-week spreadsheet exercise. By the time the forecast arrived, the picture had already changed. Too little compute could delay engineering work; too much meant hardware sitting idle. The challenge was a shared model that could connect individual chip programs to organization-wide investment decisions.
The work behind
the outcome.
Bring the roadmap into the forecast.
I built Python pipelines around LSF job logs and milestones and partnered with data science on forecasting features. The engine combined time-series extrapolation with project-milestone changes.
Let leaders ask “what if?”
I built Java/Spring Boot simulation services for concurrent roadmap scenarios. Executive mockups and integration work with engineering divisions helped validate that those scenarios supported real planning decisions.
How the pieces connect.
01LSF logs + milestones
02Python data pipelines
03Recursive demand model
04Forecast + scenarios
05Investment planning
Explanatory system view · reconstructed from the project
Hands on.
People with me.
I owned product definition, the backlog, and the data-model specification while building pipelines and simulation services. I partnered with data science, engineering divisions, and executive users to take the platform through launch.
PythonJava / Spring BootMySQLTime-series forecastingScenario simulation
What came out of it.
The platform supported approximately $4M in annual infrastructure savings and approximately 90% forecast accuracy over a two-year horizon. Capital planning fell from four weeks to two. These are platform outcomes achieved through the wider team’s work.
CONTEXT & LIMITS
The platform and its underlying business data are internal. The diagram is an explanatory reconstruction, not a production screenshot or a disclosure of forecast data.
Before forecasting: make the infrastructure move.
My broader Qualcomm work ran from January 2023 to August 2025. Earlier, I built migration workflow services, Python dependency checks, and ServiceNow integrations across 12+ teams. That separate platform delivered 40% faster transitions and 60% fewer mid-transition failures. As an intern, I also worked on developer automation and a migration of 50+ containerized applications.
Coupang
AI is running. Is it working?