Why We Invested in Firmus

Mike Heller
Navitas Capital
Published in
4 min readMay 29, 2024


AI-enabled risk management for construction design

Looking back through time, design has always been one of the first steps in creating pretty much any man-made structure, from the ancient pyramids to modern skyscrapers. Every successful construction project relies on a high-quality design plan. It doesn’t matter if you are building a log cabin in the woods or a basketball arena downtown; a strong design has a massive impact on the project’s success.

In modern construction, multiple project stakeholders collaborate to create the design plans — architects, engineers, developers, contractors, the list goes on. The ‘design phase’ is typically the precursor to a broader ‘preconstruction process,’ because a project’s design directly informs how much the project will cost. To effectively bid on a project, contractors must know how many materials they need to buy, how long the project will take, how complicated the construction methods are, among a host of other considerations. The answers to most of these questions can be found in the project’s design.

Bleary eyes

But design documents are complicated! They often include hundreds, if not thousands, of lines and grids and objects and annotations. Luckily, construction professionals are expert decoders and can derive key insights when reviewing designs. However, poring over pages and pages of complex documents takes countless hours and can cause even the most focused professional’s eyes to blur. Sometimes contractors conduct this painstaking design review before they even bid on a project. So all this time and effort is spent on a project that might not even lead to revenue.

Source: Fiverr

To complicate matters further, designs are rarely, if ever, 100% accurate and error-free. Anytime you have multiple stakeholders collaborating on a document undergoing countless iterations, there are bound to be some gaps. In many disciplines, documents that are nearly complete are good enough. This is not true in construction. Issues in designs can lead to massive increases in costs and project delays. In fact, experienced builders often highlight missing and/or incorrect information in design plans as the leading driver of cost and schedule overruns when preparing project post-mortems. These issues can account for up to 5–8% of a project’s total costs!


Given all this complexity and the importance of total accuracy, how can we better review and refine designs to streamline construction? Well, Firmus has found a way to apply best-in-class computer vision and artificial intelligence technology to help general contractors better manage design risk. Users simply upload PDF designs, and the machine goes to work analyzing the plans. Firmus then automatically produces a list of potential issues that may require resolution along with an annotated set of plans. Doors without identifiers? Conflicting locations and information between design disciplines? Objects and materials missing from the design schedule? Firmus has you covered. For contractors, this process is incredibly simple and creates tremendous value. Gone are the days spent tracking 500+ door annotations or meticulously cross-referencing objects and text across multiple pages. Instead, Firmus delivers accurate, reliable design reviews in a fraction of the time.

The Firmus team brings tremendous technical talent to bear in creating products to alleviate construction design pain points. Namely, Shir Abecasis (CEO) has extensive experience leading technology teams within the elite 8200 intelligence unit of the Israeli military. Automated design review is a hard problem to solve. Existing solutions offer an approach grounded in workflow software that enhances collaboration but does not automate the review process. By leveraging advances in machine learning, applying their unique technical capabilities, and building a world-class technology team, we believe the Firmus team is poised to deliver next-level automation to design review.

Where’s the BIM?

Those familiar with design software tools might ask… But what about 3D design (aka BIM — building information modeling)? Isn’t BIM the future of construction design? BIM is an incredible innovation in the world of architecture, engineering, and construction (AEC). Structures are built in three dimensions, so it makes intuitive sense for them to be designed in three dimensions as well. Despite the overwhelming promise of this technology, most construction projects still rely on some form of 2D plans to guide the building process. In practice, 3D models are often converted to 2D plans to incorporate additional details that are challenging to model. Sometimes projects rely on 2D plans to facilitate collaboration among project stakeholders that might be less comfortable with BIM. In the end, often 2D plans contain more details applicable to the building process (albeit less structured) than the initial 3D models.

Here is where it gets really exciting. Through novel deep learning techniques, Firmus can contextualize all the information from 2D designs into a searchable database. Leveraging this data set, Firmus will be able to answer any question related to project design. How many windows are there on the 3rd floor? Where is there exposed ductwork? All questions that can be answered via such a searchable database. Firmus is effectively reverse engineering a more detailed and accurate building information model (BIM) with the detailed data from 2D plans — the documents typically used during the build.

Onwards and upwards

Firmus will unite design disciplines and contractors within a comprehensive platform to better manage design risk, enabling faster builds, less rework, and more informed bids. As a true AI agent, Firmus will continue to push the bounds of impactful automation that solves true business problems within the preconstruction process.



Mike Heller
Navitas Capital

LA-based investor in technologies creating value for real estate and construction