Last Updated: 1 July 2026
Every building with sprinklers needs a plan. Where each head goes. How the pipes connect. Whether, if a fire actually starts, the system puts it out fast enough. None of that gets approved on a guess, because the safety codes don't leave room for one. A mistake on paper can mean a system that fails when it matters most. So engineers draw it by hand and run the math by hand, project after project, even though most buildings need roughly the same things. That slowness has a cost: fire protection companies turn down jobs they don't have the staff for, while demand keeps growing faster than the industry can train people to meet it.
Jason Tielve spent twenty years inside that slowdown before he tried to fix it. He ran a company that installed and serviced sprinkler systems, so he wasn't guessing at the problem. He watched skilled engineers spend hours on work that looked almost identical from one project to the next, which eventually led him to build FireDesign.ai alongside Omar Hafez, founder of the software company Think Big Technology and now FireDesign.ai's Head of AI Innovation. Omar brought what Jason didn't have: years building AI products, aimed at a problem nobody in the industry had solved well.
A user uploads a building's floor plan, and the software figures out where sprinklers need to go, checks the layout against fire safety rules, and helps plan how the pipes connect, all in a fraction of the time it would take by hand. A licensed engineer still has to review and approve everything before it's used. Nothing skips that step. The goal isn't to remove the engineer. It's to get them to review faster, with the groundwork already done. Looking further out, the company wants this same approach to eventually cover other parts of building design too.
That's worth paying attention to: few industries combine "this needs to be fast" with "this can't afford to be wrong" quite like fire protection. Jason and Omar explain how the AI actually works, where it stops and a person takes over, and what it took to earn the trust of a notoriously cautious industry.
1. For readers meeting FireDesign.ai for the first time: what does the platform actually do, and who is it for?
Jason Tielve
FireDesign.ai is an AI-powered platform that automates the design of fire sprinkler systems. A contractor or engineer can upload a set of architectural plans, and the platform generates a code-compliant sprinkler layout in a fraction of the time it traditionally takes.
We built it for fire protection contractors, consulting engineers, designers, and engineering firms, really anyone responsible for producing fire sprinkler designs. Whether you're a small shop trying to keep up with demand or a larger organization managing hundreds of projects, our goal is the same: help you complete quality designs faster without compromising the engineering judgment that makes this profession so important.
At the end of the day, we're not trying to replace designers. We're giving them a tool that eliminates repetitive work so they can focus on the decisions that truly require their expertise.
2. You founded FireDesign.ai after twenty-plus years in fire protection. What's that story, and why did software end up being the answer?
Jason Tielve
I've spent more than twenty years in the fire protection industry, building and leading a company that grew into a business generating over $20 million. During that time, I was involved in virtually every part of the process, from estimating and design to installation, inspections, and managing large teams.
One thing never changed: sprinkler design was always the bottleneck.
The demand for fire protection continues to grow, but the number of experienced designers isn't growing at the same pace. We were constantly asking talented people to spend countless hours on repetitive drafting work instead of applying their engineering knowledge where it mattered most.
After seeing that challenge year after year I realized the industry didn't just need more designers, it needed better tools.
That's really how FireDesign.ai began. It wasn't about building an AI company. It was about solving a problem I'd lived with for decades. We started asking, "What if technology could handle the repetitive portions of design while leaving the critical engineering decisions in the hands of experienced professionals?" Everything we've built has been centered around answering that question.
3. You run Think Big Technology and serve as Head of AI Innovation here. How did you end up working on fire protection specifically, and what does that split between your two roles actually look like day to day?
Omar Hafez
There has always been a family connection. Jason's business partner at 24/7 Fire Protection is my uncle, so we've known each other for years. As Think Big Technology continued growing in AI and software development, and Jason built one of the industry's leading fire protection companies, it became clear we could combine our strengths to solve a long-standing problem in fire sprinkler design.
Today, I split my time between leading Think Big Technology and serving as Head of AI Innovation at Firedesign.ai, where I focus on AI strategy, product direction, and working with our engineering team and industry experts to build practical solutions for the life & safety domain.
4. What does fire sprinkler design actually look like today, without a tool like this? Why does it take so long?
Jason Tielve
Traditional sprinkler design is an incredibly detailed and manual process. A designer begins by reviewing architectural drawings, identifying hazards, determining occupancy classifications, laying out sprinkler heads, routing pipe, coordinating around structural and mechanical systems, and ensuring everything complies with NFPA standards and local requirements.
Every one of those decisions affects the next. Even a relatively straightforward building can involve hundreds or thousands of individual design decisions.
The work takes time because accuracy matters. Designers aren't just drawing lines, they're protecting lives and property. They also spend a tremendous amount of time performing repetitive tasks that don't necessarily require years of experience but still consume hours of their day.
As projects become more complex and labor becomes harder to find, those timelines continue to grow. That's where automation can make a meaningful difference by handling the repetitive portions of the workflow while allowing engineers to focus on reviewing, refining, and validating the design.
5. Why does this particular stage, turning architectural plans into a sprinkler layout, lend itself to automation when so much of fire protection engineering doesn't?
Omar Hafez
This stage is especially automatable because it sits in a rare “sweet spot” of structure and rules.
Architectural plans already contain clear geometry—walls, rooms, ceilings—so the system has something consistent to read from. Once that’s extracted, sprinkler layout becomes a constraint-solving problem: spacing rules, coverage radius, obstruction rules, and code requirements (like NFPA standards) are largely deterministic and repeatable.
That combination makes it less about subjective engineering judgment and more about pattern + rules + optimization, which AI handles well.
In contrast, much of fire protection engineering involves real-world ambiguity—field conditions, installer decisions, clashes with other trades, and inspector interpretation—which are harder to standardize or automate reliably.
6. Walk through what happens between the moment someone uploads an architectural plan and the moment FireDesign.ai produces a sprinkler design. Where does the AI make the calls, and where do fixed engineering rules take over instead?
Omar Hafez
When a user uploads an architectural plan, our AI first analyzes the drawings to identify walls, rooms, dimensions, and other relevant building features while extracting the information needed for design by using our proprietary geometry extraction API. From there, it classifies the project, understands the building layout, and prepares the data for the engineering engine.
Once the building has been interpreted, deterministic engineering rules take over. Our rules engine applies the applicable NFPA requirements, calculates sprinkler placement and spacing, validates code compliance, and generates the final design. In short, AI handles understanding the plans and making sense of unstructured data, while the engineering logic ensures the final output follows established fire protection standards.
7. Reading a plan means resolving units, classifying layers, grouping floors, extracting geometry, identifying rooms, and spotting doors, windows, and obstructions, with AI stepping in when something's unclear. How does the system decide a plan has been read well enough to move on to actual design work?
Omar Hafez
The system doesn't move forward simply because it found enough information—it moves forward once it reaches a confidence threshold and all of the required design elements have been validated. We combine AI confidence scores with deterministic validation checks to ensure key information such as scale, geometry, rooms, and other critical building features have been accurately identified before design begins.
Before the design process starts, users are presented with a preview of the interpreted plan where they can review the extracted geometry, room boundaries, and other detected elements. This manual review stage allows users to confirm the AI's interpretation or make corrections if needed. If anything is missing, inconsistent, or falls below our confidence threshold, the system flags it for additional processing or user review rather than guessing. In a code-driven industry like fire protection, it's better to verify than make assumptions that could impact the final design.
8. In your experience, what kinds of drawings give the system the most trouble? What does a genuinely messy or unusual plan look like in the real world?
Jason Tielve
Anyone who's been in this industry knows that not every set of plans is clean. You see incomplete architectural drawings, multiple revisions layered on top of each other, handwritten markups, missing dimensions, unusual ceiling conditions, renovations to older buildings, or projects where several disciplines haven't fully coordinated yet.
Those are the projects that challenge both people and software.
Our philosophy has never been to pretend every drawing is perfect. Instead, we've spent a great deal of time teaching the platform to recognize uncertainty. If something isn't clear, we'd rather flag it for review than make assumptions.
The reality is that experienced designers already know when a project requires additional judgment, and we believe software should support that judgment, not replace it. Some projects are simply more complex than others, and that's exactly where collaboration between AI and experienced professionals becomes most valuable.
9. At a high level, what's actually inside the AI stack? Computer vision, geometric reasoning, optimization, language models, rules engines, some mix of all of it?
Omar Hafez
Conceptually, AI is used to interpret & classify the geometry, extract context, and understand what is being depicted in the drawing. Next, geometric reasoning works to spatially map rooms, walls, boundaries, dimensions, and designable area.
Afterwards, it stitches together an optimization algorithm with a deterministic rules engine. The optimization calculates smart sprinkler layouts and the rules engine throws at it the relevant NFPA requirements and design constraints. The AI is not flying by itself—there are hard-set engineering rules that guide the outputs to be both smart and code-specific.
10. The system is described as grounded in verified codes and professionally reviewed plans. How were those source materials chosen, labeled, and turned into something the AI could actually learn from?
Omar Hafez
It wasn’t our approach to “feed it lots of data and hope it learned the correct patterns.” We partnered with Jason’s team of fire protection experts to source and use high-quality training data – confirmed plans, code-compliant designs matched with their corresponding NFPA standards. We then cleaned, organized and labeled that data to train the system to recognize building features, extract design intent, and know how architectural attributes relate to fire protection design requirements.
Equally important, we delineate what AI learns vs. what it is allowed to determine. AI learns how to read plans and interpret information while compliance determinations are dictated by a deterministic rules engine built from validated engineering principles and tested continually through expert review. This hybrid approach allows for consistency, transparency and code compliance.
11. When FireDesign.ai lays out a pipe route, what matters most to you as someone who's actually had to install these systems? Code compliance, ease of installation, material costs, something else?
Jason Tielve
All of those factors matter because they're interconnected, but if I had to prioritize them, code compliance comes first. Nothing else matters if the system doesn't meet code.
From there, I think like a contractor because that's where I spent much of my career. A design also needs to be practical to install. It should minimize unnecessary complexity, reduce labor where possible, coordinate well with other trades, and make efficient use of materials without sacrificing performance.
One of the advantages of having an installation background is that you understand the difference between something that looks good on paper and something that crews can actually build efficiently in the field.
That's the mindset we've tried to bring into FireDesign.ai. We're not just optimizing for drawings; we're optimizing for real-world construction.
12. Does AI play a direct role in hydraulic analysis, or does that step run on standard engineering calculations and outside software instead?
Omar Hafez
AI currently does not run hydraulic calculations. AI creates the optimized code compliant sprinkler layout and hydraulic information to run calculations. Calculations are run using proven engineering practices and standard industry software. This allows for calculations to be done in a way that is in line with accepted engineering practices and can be reviewed, verified and stamped by a licensed engineer.
AI could be used in the future to help engineers see where they can optimize or may potentially have hydraulic issues earlier in the design process. The calculations themselves would still be based off deterministic engineering not predicted by AI.
13. What does the evaluation process look like before a new model or AI feature gets added to a live FireDesign.ai workflow?
Omar Hafez
Every new AI model or feature release goes through stages of validation before deployment to production. We run it against handpicked benchmark projects, compare the output to code-compliant designs by our team of fire protection engineers, and score its accuracy and consistency. The engine must also pass deterministic validation rules ensuring no conflict with our engineering rules.
Only after demonstrating it can reliably meet our performance expectations does it enter our live workflow. Because this is a life safety application, when we release a new AI capability we don't just want it to be accurate, we need to know it will perform reliably and predictably before our customers ever use it.
14. The structured engineering report includes confidence scoring. What does that score actually measure, and does it change what gets processed automatically versus sent off for manual review?
Omar Hafez
The confidence score measures how certain the system is that it has correctly interpreted the architectural plan—not whether the engineering design itself is correct. It takes into account factors such as drawing quality, geometry extraction, object recognition, layer classification, and the completeness of the information needed to generate a design.
That score directly influences the workflow. Plans with high confidence move through the automated design process, while plans with lower confidence or missing information are flagged for manual review. Users also have an opportunity to review and confirm the interpreted plan before the system proceeds, ensuring the AI's understanding is validated before any engineering design is generated.
15. NFPA 13, 13R, and 13D rules are built directly into the platform, and compliance checks default to fail-closed. How is that regulatory logic actually divided up among hard-coded rules, AI components, and standard calculations?
Omar Hafez
There's a deliberate division of responsibility when it comes to logic. AI is responsible for reading plans, extracting building information, and disambiguating plan elements, but making code compliance decisions is handled separately. NFPA requirements are baked into deterministic rules that explicitly control where sprinklers are placed, their spacing & coverage, and other design limitations. Standard engineering calculations are used whenever math can verify validity.
The system is also designed fail-closed. If information is missing, conflicted, or unable to be validated the system won't guess or autoflow. Ambiguity is flagged for review (or further processing) by the user. Code compliance is never left up to AI.
16. Every validation is said to trace back to a published standard. When the platform places a sprinkler or flags a compliance issue, what exactly can the reviewing engineer actually look at?
Omar Hafez
Every design decision is fully traceable. When the platform places a sprinkler or flags a compliance issue, the reviewing engineer can see the relevant design parameters, the validation results, and the specific NFPA code references that were applied. This provides a clear explanation of why a decision was made rather than treating the AI as a black box.
That level of transparency is critical because the platform is designed to assist engineers, not replace them. Every output can be reviewed, verified, and, if necessary, modified before the final design is approved and sealed by a licensed professional.
17. Licensed engineer review is still required for certification, site-specific water supply decisions, and approval from the authority having jurisdiction. Beyond that, what else do you check personally, and are there decisions you believe should never be handed off to the platform, full stop?
Jason Tielve
Absolutely. AI should assist professionals, not replace professional responsibility.
Even after a design is generated, I want experienced engineers reviewing the overall strategy, verifying coordination with other building systems, evaluating constructability, considering project-specific conditions, and applying the judgment that only comes from years of experience.
Every building has unique circumstances. Water supply conditions, special hazards, owner requirements, local amendments, and coordination with the authority having jurisdiction all require human expertise.
I don't believe software should ever become the final decision-maker for life safety. Those decisions belong to licensed professionals who understand the project as a whole and are ultimately accountable for the final design.
Our goal has always been to make engineers more productive, not less essential.
18. How does an engineer's day-to-day work actually change once they start using this platform? Which repetitive tasks shrink or disappear, and what new kinds of review does the system hand them instead?
Jason Tielve
The biggest shift is that engineers spend less time creating the initial design and more time evaluating it.
Instead of manually placing every sprinkler head or routing every section of pipe from scratch, they begin with a generated design that can be reviewed, refined, and optimized. That means they're spending more of their day making engineering decisions instead of performing repetitive drafting tasks.
The review process also becomes more intentional. Engineers are looking at constructability, project-specific challenges, coordination with other trades, and ensuring the design aligns with the unique requirements of that building.
That's where I believe experienced professionals create the most value. Technology can accelerate the work, but engineering judgment is still what delivers a safe, efficient, and reliable fire protection system.
19. Beyond the design-time savings that have come up in past coverage, what real-world outcomes are actually being tracked, or planned to be tracked? Things like override rates, correction rates, or review time?
Omar Hafez
Of course we're interested in reporting design- time savings, but beyond that we care about measurements that matter to actual engineering workflows. Manual override rates, rates of change during review, trends in confidence scores, review time, rate of plans needing additional user input prior to design – these are examples of measurements that will help us identify where the AI shines and where there is opportunity for continued improvement.
When we start hitting critical mass on adoption, we will also be looking at design turnaround time, standardization across projects, and decreased repetitive engineer hours. We're not looking to simply spit out designs as fast as possible. We want to decrease review effort without sacrificing the expected accuracy and code compliance necessary for a life safety application.
20. The company's stated long-term goal is an AI-native engineering platform that eventually covers more than just fire sprinklers. What does that roadmap actually look like, and why start with fire protection?
Jason Tielve and Omar Hafez
We started with fire protection because it's a highly specialized discipline with well-defined engineering standards, significant labor shortages, and a design process that benefits tremendously from automation. It also gave us the opportunity to work alongside people that Jason has worked with for his whole professional lifespan to solve a real problem before expanding into adjacent trades.
Long term, our vision is to build an AI-native engineering platform that supports multiple building systems, including plumbing, mechanical, electrical, and other MEP disciplines. Each trade will follow the same philosophy: use AI to understand the building and automate repetitive tasks, while deterministic engineering logic and applicable codes ensure every design remains accurate, transparent, and compliant.