Inspiration

The idea for RENNEL came from noticing a real gap. When we looked at software built for mission critical construction, hospitals, data centers, government facilities, anything where a failure isn't just expensive but genuinely dangerous, there wasn't much out there built specifically for that world. Most tools treat every project the same way, whether it's a strip mall or a facility that has to survive an earthquake without losing power. We felt that mission critical work deserved something built around its actual stakes: real time checking, deterministic proof, and no room for a system that's only right most of the time.

What it does

RENNEL takes raw design and project data and automatically checks it for compliance, risk, and cost issues as it's being built, instead of waiting for a manual review weeks later. It flags problems the moment they appear, points to the exact rule or clause that was violated, and shows the calculation behind that flag so nothing feels like a guess. A simple dashboard pulls all of this together into one place, showing a project's compliance score, budget status, and open risks at a glance.

How we built it

We started by designing a shared way of describing project information, so that a wall, a slab, or a line item could all be represented the same way, along with how each one relates to everything around it. On top of that, we built a small pipeline where an AI layer reads through this information and figures out which rule applies, and a separate, fixed calculation engine actually runs the check and produces the result. That kept the AI focused on understanding and routing rather than making the final call. We then added a lightweight dashboard on top so teams could see flags, costs, and risk without needing specialist software, and made sure the whole thing could run on local infrastructure so sensitive project data stays private.

Challenges we ran into

The hardest part was building a shared format general enough to describe very different kinds of information, like a structural element and a contract clause, without making it so abstract that it stopped being useful. We rebuilt this a few times before it felt right. We also had to resist the temptation to let the AI shortcut straight to an answer, since early versions occasionally invented a rule that didn't actually exist. Separating understanding from calculation fixed this, but it took a fair amount of trial and error to get the split right.

Accomplishments that we're proud of

We're proud that every flag RENNEL produces can be traced back to a real rule and a real calculation, rather than being a plausible sounding guess from a language model. We're also proud that the same underlying approach we built for construction compliance turned out to extend naturally to procurement, catching budget and contract issues with the same logic we used for structural and code checks, without having to rebuild the system from scratch.

What we learned

The biggest thing we learned is where AI actually earns its place in a high stakes system. It's not at the end making the final judgment call, it's earlier, doing the understanding and translation work a person would otherwise do by hand, while the actual decision comes from fixed, checkable logic. We also learned how much more powerful information becomes once you represent it by its relationships to everything around it, rather than just as an isolated fact.

What's next for RENNEL

The next step is to build the smallest working version of the full idea: one jurisdiction, one mission critical building type, and one real project model, run end to end so we can prove the approach holds up outside of a demo. From there, we want to expand the same engine further into procurement workflows, testing it against real contracts, purchase orders, and vendor data to see how well it catches issues before they become costly.