Solar energy
production
SaaS simulation tool
Client
PVcase
Involment
Senior Product designer
Skills used
User Research & Analysis
Information Architecture Complex Data Visualization Rapid Prototyping
QA & Validation
Tools used
Figma
Claude design
Gemini AI
Dovetail
Jira
Mix panel
Date
2025 – 2026
Client
Hewlett-Packard
Date
2019 – 2021
Involment
UI designer – Accessibility specialist
Skills used
Interaction design – User test
Tools used
Axure – Jira – Sketch
Context
PVcase Yield is an advanced solar energy simulation platform used by engineers, developers, and financiers to evaluate and optimize the energy production of multi-million-dollar utility-scale solar projects.
Operating as the sole Senior Product Designer within an agile cross-functional squad (1 PM, 1 Lead Engineer, 3 Frontend Developers, 3 Backend Developers, and 3 Domain Scientists), I owned the full design lifecycle from raw research to final visual specifications and QA.
The core challenge was consistent across the entire product: How do you translate highly complex scientific equations and technical inputs into intuitive, decision-grade user interfaces without losing precision?
Over a 15-month period, I led the UX/UI strategy for three major feature releases that transformed how engineers assess uncertainty, simulate equipment limits, and model energy losses.
Probability estimation
The challenge
Solar energy estimates are inherently uncertain. Financial institutions and project developers require statistical probability outputs (P50, P75, P90, P95, P99) to secure bank approvals and manage risk. Historically, engineers were used to single-number outputs. Introducing a full suite of uncertainty variables threatened to cause information overload and disrupt the existing navigation.
Research & discovery
I conducted user interviews to uncover the exact context of this data.
The “Why”: Users didn’t just want numbers; they needed to understand the financial risk profile of the plant to adjust the layout if energy production fell short of project thresholds.
The Artifact: They needed a highly professional, definitive document to present to banks and stakeholders for project approvals.
Design approach
I restructured the navigation to introduce a dedicated uncertainty section that allows progressive disclosure. Users can fine-tune the precision of their estimations depending on the project phase. I focused heavily on data density, ensuring the P-values were highly scannable alongside standard yield metrics. I also designed a clean, formal PDF export engine so engineers could immediately download a bank-ready report to drive stakeholder discussions.
Results
- Support tickets related to probability estimation dropped significantly in the months following launch
- Feature adoption tracked via PDF export: users actively including probability values in client and investor documentation
- Validated positively in follow-up sessions — users reported that the output gave them what they needed to have conversations with banks and clients
Inverter temperature derating
The challenge
Inverters lose efficiency as they heat up—a concept known as temperature derating. To simulate this accurately, we needed to introduce seven new technical data fields and a corresponding performance curve graph into an already highly dense, text-heavy equipment modal window.
Discovery
Because this was a highly specialized, physics-driven feature, we skipped traditional user interviews. Instead, I ran an internal discovery process with our 3 embedded domain scientists. We analyzed competitive tools and deep-tech industry standards to understand how engineers visually conceptualize derating curves.
Design approach
The design problem was purely about screen real estate, visual hierarchy, and progressive disclosure.
I structured the 7 new input fields logically to mirror the engineer’s mental workflow.
I custom-built a clean, interactive vector graph directly into the modal. The graph updates dynamically as the inputs change, letting engineers catch data-entry errors instantly before running a time-consuming simulation.
Working within a legacy design system, I hand-crafted bespoke iconography and UI components to fit perfectly into the existing interface layout.
Impact & validation
Frictionless Launch: Customer support tickets regarding inverter configuration ceased entirely after launch; the visual graph successfully eliminated user confusion regarding input parameters.
MV/HV transformers
The challenge
Previously, the Yield simulation engine treated electrical transformers purely as an abstract “loss percentage” rather than physical equipment components. This forced a massive limitation: users had zero control over real-world transformer parameters, causing inaccuracies in large-scale Medium Voltage (MV) and High Voltage (HV) setups.
Changing this meant completely restructuring the entire loss chain across multiple system sections: Layouts, Losses, Results, and PDF exports.
Discovery
I conducted a deep round of user interviews to understand how engineers map electrical infrastructure in their minds. To accelerate the timeline, I utilized Gemini AI to brainstorm user archetypes and analyze large amounts of qualitative interview text inside Dovetail.
Critical Insight Uncovered: Engineers definitively perceive transformers as physical components (just like solar modules or inverters), not abstract calculations. Furthermore, they demanded granular control over both the MV transformer/cabling and the HV substation transformer/cabling.
Design approach & prototyping
To validate a solution quickly, I built a functional, data-driven prototype. Because scalability was key, I discovered that users rarely configure more than one MV and one HV transformer type per site.
Instead of complex multi-column grids that would crowd the screen, I designed a streamlined, single-column infrastructure panel.
This allowed users to easily toggle transformers on/off and inputs shifted fluidly across Layout, Loss, and Result screens to match their physical mental model.
Impact & validation
Feature adoption: Product telemetry revealed that 60% of all platform simulations now run with both MV and HV physical transformers actively configured.
Platform scalability: Successfully shifted the tool’s core architecture from abstract calculations to an object-oriented, component-based layout model.
Conclusions
Good UX design in complex SaaS is not about simplifying the domain; it is about clarifying the complexity. By respecting the deep technical expertise of the user, leveraging AI tools to speed-up research synthesis, and acting as a collaborative bridge between engineering, data science, and product management, I was able to deliver interfaces that turned complex scientific computations into clear, actionable business strategies.