Case study · 2022 — 2025
Solar Analysis Platform
Shade analysis and PDF reporting for a solar platform, with the heavy computation cut from minutes to milliseconds.
- Outcome
- 10 min → 10 msShade analysis computation8 min → 4 minPDF report generation
- Role
- Fullstack Developer, later Backend Specialist
- Context
- SOLO (Utah, US), while at AssureSoft
- Stack
- PythonNumPyGCPKubernetesPuppeteerReact
- Links
- Ask about this work
Summary
SOLO's platform runs shade analysis and produces PDF reports for solar projects. I started as a fullstack developer, optimizing business logic, resource allocation and PDF generation, then moved to backend specialist, where I rewrote the shade-analysis computation and the projection algorithms behind it.
The problem
Two steps gated every project: a shade-analysis computation that took around ten minutes, and PDF report generation that took eight. Running at scale on Google Kubernetes Engine, both were compute cost and waiting time for every user.
Constraints
Results had to stay numerically faithful: a faster projection that disagreed with the old one was worse than a slow one.
The work landed inside a live product, release by release, without a rewrite.
What I did
Cut PDF report generation from 8 minutes to 4 with Puppeteer on GCP.
Rewrote the shade-analysis computation with Python and NumPy, from 10 minutes to 10 milliseconds, a 99.998% reduction that cut computational cost on the Kubernetes clusters.
Improved the projection algorithms with linear algebra and accelerated the surrounding data processing.
Fostered clean code and TDD through code reviews and mentorship.
Client name cleared for publication; internal system details are not disclosed.