Product 8 min read

Geom Left Stealth With Four Named Homebuilders and a $5M Seed. The Target Is Drafting, Not Design.

Shanna Tellerman founded Modsy and was Lennar's chief product officer before starting Geom. The company launched on 18 September with Pulte, Century Communities, Meritage and True Homes as collaborators and a deliberately unglamorous pitch: automate the weeks between a finished design and a permit-ready set.

A hand-drawn architectural blueprint showing a cross-section through a multi-storey house, with staircases, floor levels and window openings drawn in white line on blue paper.

Geom came out of stealth on 18 September with a product for production homebuilders and a list of collaborators that is the most interesting thing in the announcement. Per the company’s press release, dated from San Francisco, Geom “today emerged from stealth, collaborating with more than half a dozen of the nation’s largest production homebuilders,” and names four of them: Pulte, Century Communities, Meritage and True Homes.

The company was “founded earlier this year with $5 million in seed funding led by Elad Gil & Co and a group of prominent Silicon Valley investors,” the release says. Geom’s about page lists the backers as “Elad Gil & Co, Home Technology Ventures, Verified Capital, BBG, Aaron Levie, Guillermo Rauch, Eric Wu, and other prominent technology investors.”

The Problem Is the Back Half of the Drawing Set

Much of the AI attention in residential design has gone to the front of the process — generating layouts, exploring massing, producing renderings. Geom is aimed at the part afterwards, and its founder is blunt about where the time actually goes.

“Today, the design phase is actually relatively quick,” Shanna Tellerman told Business of Home in an interview published 17 September. “Getting the detailing where everything is dimensioned and annotated. … You have electrical, you have foundation plans. … That’s what takes weeks, sometimes even months. Those parts, in our mind, can be entirely automated.”

That is a specific and testable claim about where cost sits in production homebuilding, and it points at a category of work that is genuinely enormous and almost entirely invisible from outside the industry. A national builder does not design a house once. It designs a plan, then produces a lot-specific document set for every instance of that plan — mirrored, re-sited, re-optioned, adjusted for division standards and local code, and reissued every time something upstream changes. The design is a one-off. The drafting is a production line.

Geom’s own framing, per its website, is “Your plans, understood. Your drafting, automated,” with output described as “Lot-level and community-level CDs in DWG and PDF.” The release positions the product as fitting the tools builders already run rather than replacing them.

“The industry has been asked to do more with the same 40-year-old tools for far too long,” Tellerman said in the release. “We are building Geom to fit into the workflows builders already use and automate the tedious work, so their teams can focus on the designs and decisions that deserve their expertise and judgment.”

The Two Numbers

The release gives two worked examples, and they are the company’s own figures, reported as the company reports them:

“In one project, Geom reduced plan mirroring from weeks to minutes, cutting the cost of the workflow by 73%. In another, Geom reorganized master sheet sets into construction-document sets for more than 300 plans in 20 minutes, a process that previously required two to three days and was vulnerable to manual error.”

Plan mirroring is worth explaining, because it is the clearest illustration of why this problem exists at all. When a builder needs the reverse-handed version of a plan — the same house flipped to suit the lot next door — the geometry flips trivially and everything attached to it does not. Text has to stay readable, dimension strings have to be re-anchored, door and window schedules have to be re-mapped, and anything asymmetric in the mechanical or structural layout has to be reconsidered rather than reflected. It is a task that is conceptually trivial and practically expensive, repeated across thousands of lots, and it is exactly the shape of work that a system with semantic understanding of a plan set should be able to take.

We have not independently verified the 73% figure or the 300-plan example, and the release does not name the projects, the builders involved or the baseline cost.

The release also sizes the opportunity: “In a market that spends $400B annually on AEC services, production homebuilding is Geom’s starting point.” That is the company’s own framing of its addressable market, not an independent estimate.

Who Is Building It

Tellerman’s background is the reason the collaborator list is plausible. Business of Home reported on 31 August 2023 that Lennar had acquired Modsy, and that “Modsy founder and CEO Shanna Tellerman will become Lennar’s chief product officer.” In this month’s interview, BOH reports that following Modsy’s 2022 shutdown she “spent three years at Lennar, the homebuilding giant that acquired Modsy’s IP,” developing technology for the company as chief product officer.

Geom’s about page adds that before Modsy she “founded Wild Pockets (Sim Ops Studios), a Carnegie Mellon University spinout acquired by Autodesk in 2010,” and describes the company as “a team of architects, homebuilding operators, and AI engineers.”

The timeline is tight. Per BOH, Tellerman hired a “sub 10” team and “officially got to work in April of this year”; five months later the debut product is launching. Her stated reason for starting with production builders rather than custom homes is a volume argument: “Working with dozens of production homebuilders is going to have a universal impact faster than working on lots of one-off custom homes.”

The one builder-side voice in the release is Kieran Hardesty, whom the release identifies as VP of National Architecture at Century Communities: “Geom is helping streamline some of the most labor-intensive and time-consuming processes in residential architecture through the use of AI. Century Communities is excited to collaborate with their team as they work to improve how builders manage plans, reduce manual effort, and produce construction documents more efficiently.”

Read that quote closely and it is an endorsement of direction rather than a deployment claim — “as they work to improve” is the operative phrase. The release is consistent with itself on this point: it describes the relationships as collaborations throughout, and it also states, in its own words, “Initial pilot workflows have produced measurable results.” This article uses the release’s vocabulary for the same reason.

Analysis: The Same Investors Keep Betting on the Same Thesis

Analysis. Two of the names on Geom’s investor list also turned up in this newsroom’s recent coverage. Home Technology Ventures appears among the investors in Scaffold’s $15M seed, dated 15 September, and Eric Wu is the Opendoor co-founder behind NavigateAI, which came out of stealth on 26 May with a $25M seed our reporting recorded as led by Elad Gil. Geom’s release names the lead here as Elad Gil & Co.

That is a small group of investors placing repeated bets on the same proposition: that residential construction’s software problem is not a missing application but unpriced manual labour inside workflows that already exist. Scaffold’s co-founder and CEO Ben Johnson put it as builders’ systems needing “to stop disagreeing with each other.” Geom is putting it as drafting.

The residential side of the market has now produced several well-funded attempts at overlapping territory — Higharc raised $95M for a spatial-database approach that carries a builder’s plan automatically through estimates, construction documents and marketing renderings, and Digs raised $25.3M in a Series A led by Builders FirstSource. Geom’s stated differentiator is that it does not ask a builder to move onto a new platform: the about page’s framing is that production homebuilding “runs on real-world complexity: national standards, regional and division variation, sales option workflows, and product catalogs,” and the product’s claim is to read that complexity out of the builder’s existing files rather than require it to be re-entered.

Whether that is a durable moat or a temporary integration advantage is the open question, and five months of operation is not enough to answer it.

What Is Not Established Here

We scanned the release twice, in differently worded passes, for commercial terms. The words “price,” “pricing,” “customers,” “client,” “paying” and “contract” do not appear in it; the words that do appear are “collaborating” and “pilot.” So nothing in the announcement establishes that any of the four named builders is a paying customer rather than a development partner, and this article does not say that any of them is.

The only headcount figure we have is BOH’s “sub 10.” We did not find a revenue figure, or the identity of the remaining builders implied by “more than half a dozen,” in any of the sources listed below.

The $5 million seed is described in the release as the funding the company was founded with earlier this year, not as a round being announced now. We did not find a valuation for Geom in the release, the about page, the BOH interview or in search.

Geom’s press page also links a version of the announcement hosted on Construction Dive’s press-release section; that is syndicated company material rather than independent reporting, and it is not cited here as corroboration.

Sources