🧪 Kimia — TDS Generator

Creating Technical Data Sheets in the chemical industry has always been manual and repetitive. Teams spend hours gathering product information, formatting documents, checking compliance, and making sure every specification is accurate before anything reaches a customer.

Kimia wanted to rethink that workflow with AI. Not to replace the experts, but to remove the repetitive work that slowed them down. And the vision was never limited to TDS documents alone. The idea was a platform that could generate any documentation. We just started with the TDS because it was the easiest to build first and the most valuable document for our clients.

As the sole product designer, I worked alongside the founders and engineers to turn an ambitious idea into a product customers could actually use. There was no existing product to improve, so every decision had to start from first principles.

Unlike a typical SaaS product, there weren't many references for what an AI-powered TDS generator should feel like. The product itself was evolving every week. AI capabilities changed, business requirements shifted, and technical limitations constantly shaped what the experience could be.

The hardest problems weren't visual. They were about trust and clarity: making AI feel predictable instead of magical, helping users trust generated outputs, supporting both technical and non-technical users, designing workflows around incomplete product information, and keeping the experience simple despite a genuinely complex backend.

Every screen had to balance usability with the realities of how chemical professionals actually work.

Rather than designing isolated screens, I focused on the entire workflow. Working closely with engineering and the founding team, I mapped how information entered the system, how AI transformed it into structured data, and where users needed to review or edit results before generating the final document.

The product grew into several connected experiences: product onboarding, product information management, AI-assisted TDS generation, review and editing workflows, and analytics dashboards.

Because requirements were often ambiguous, a lot of design sessions started on whiteboards instead of wireframes. We explored different interaction models before committing to polished interfaces.

Being the only designer meant wearing every hat. Beyond the interfaces, I facilitated product discussions, translated fuzzy requirements into flows, built prototypes for engineering, validated usability with stakeholders, and refined interactions during implementation. I also leaned on AI tools like Claude, ChatGPT, and v0 to explore concepts quickly. Design evolved alongside engineering instead of being handed off at the end.

The biggest design goal wasn't making AI look impressive. It was making it understandable. Instead of exposing complicated AI concepts, I designed the experience around familiar product workflows.

Users always knew what information the AI was using, what was generated automatically, what still needed human review, and where they could make edits. That transparency reduced uncertainty and built real confidence in AI-generated documents.

Along the way I established reusable components and interaction patterns, so new features could ship much faster as the product kept evolving. The TDS Generator ended up being much more than a document tool. It became the foundation for Kimia's product platform, connecting AI, product data management, onboarding, and analytics into one experience.

Designing for AI taught me that trust is the real interface. The technology can be as capable as it wants, but if users can't predict what it will do or verify what it produced, they won't rely on it. The most valuable design work here was invisible: deciding what to show, what to hide, and where a human should stay in the loop.

It also sharpened how I work in ambiguity. When the product changes weekly, pixel-perfect mockups age fast. Whiteboards, prototypes, and tight loops with engineering turned out to be worth far more than polished handoffs.

This project is still evolving, and so am I with it. But it's already the clearest proof of how I want to work: AI-powered products, complex domains, and end-to-end ownership as a solo designer.