Ansatz: From Scientist-Built Prototype to Investor-Ready Platform in Three Weeks

Rebuilt with accelerated AI-guided development — from a basic internal tool to a demo-ready platform.

Ansatz AI is a Carnegie Mellon spinout built around Hierarchical Machine Learning, a technique that helps chemistry and R&D teams model formulations from small experimental datasets, without needing in-house machine learning expertise. Its platform is used for formulation design across agrochemicals, cosmetics, household products, adhesives, and other advanced materials, and Procter & Gamble has been a development partner.

The software hadn’t caught up to the science behind it. Ansatz’s platform had been built by its own scientists, not software engineers. The underlying machine learning models were powerful, but the UI/UX did not reflect the platform’s advanced capabilities. Additionally, there was an opportunity to leverage LLMs to enhance certain aspects of the platform. So we set about to:

 

  • Modernize the UI/UX and make the platform genuinely intuitive
  • Use AI to guide the process end to end, from uploading data, to configuration, to results
  • Make the platform compelling enough for companies deciding whether to invest
  • Add the ability to create and reload projects

Ansatz needed the platform to look and feel like the sophisticated technology underneath it.

The Aboard Solution

We used deep product experience and AI acceleration to rebuild Ansatz’s platform into something modern and intuitive. The platform would turn complex chemistry data into actionable results quickly, smoothly, and with minimal user effort.

What We Delivered

AI-developed code powering a new UI/UX, an LLM integrated directly into the app to explain functionality and results in plain language, the ability to create and reload projects, and a set of video demos for investor conversations.

Most importantly, we created a totally new user experience organized around LLM-guided explanations — an integrated LLM walks users through the app, their results, and the underlying chemistry concepts as they work, rather than leaving them to interpret raw model output alone

The Technical Challenge

The hardest part wasn’t the software, it was the chemistry. The Aboard team had to understand the rules, constraints, and limitations governing Ansatz’s preconfigured formulation models before they could safely touch the interface built on top of them. LLMs helped build that domain knowledge quickly; from there, the team worked closely with Ansatz to connect the data, configurations, and models into one reliable end-to-end workflow.

Results & Impacts

Weeks, Not Months
The rebuild took about three weeks, against an estimated four to six months for a traditional approach.

Modernizing Got Cheap and Fast
Work that’s traditionally slow and bug-prone, like migrating a UI to a new stack, moved quickly and with less risk under AI-assisted development, freeing the team to spend its time on product decisions instead of migration mechanics.

Key Takeaways

AI proved especially strong on low-judgment, high-volume work here: UI/UX migration, code generation, and building the team’s own working knowledge of Ansatz’s chemistry models fast enough to move at this pace.

Want to see how Aboard could work for a project like this? Get in touch