Mach9’s $34M bet on turning laser scans into usable CAD

Mach9 has raised a $34 million Series A to build AI-powered CAD software for surveying and civil engineering. The company says nearly 100 customers have used its private beta on more than 2,500 projects, suggesting demand beyond a successful demo.

First Round Capital and Bessemer Venture Partners co-led the round, with participation from Prudence and existing investor Quiet Capital, according to TechCrunch’s October 21, 2026 report.

Founded as Mach9 Robotics in 2021, the startup initially built hardware to make visual inspections faster and cheaper. Customer conversations pushed it toward another problem: civil engineering firms could already collect detailed imagery and laser scans, but turning those measurements into usable surveys took too long.

Mach9 dropped “Robotics” from its name and moved into software. It now uses in-house AI models to produce editable geometry from reality-capture data, with tools for engineers to inspect and correct the results.

That’s a sensible application of AI. Data collection has improved faster than the workflows that turn captured data into engineering deliverables. Whether the software earns its place will depend on how much total project effort it saves, including professional review, while meeting customers’ accuracy requirements.

The expensive step after data capture

Laser scanners and imaging systems capture physical sites in detail. Civil engineers still have to identify curb lines, pavement markings, and other features, then turn them into geometry suitable for design and asset management.

Better capture hardware doesn’t automatically reduce that drafting work. For a firm that already has capable scanners, software that removes hours of tracing may be a better purchase than another device.

Mach9 co-founder Alex Baikovitz told TechCrunch that manual work in incumbent tools from companies such as Autodesk and Bentley Systems can take about a week per mile. He also described projects that take months to deliver. Those are his characterizations of existing workflows, not independent benchmarks.

Large transportation mapping assignments repeat similar extraction tasks over substantial distances. Automating some of that work could let surveying firms take on more projects without adding drafting staff at the same rate.

Mach9 says its software makes the process 10 times faster and substantially reduces costs. Buyers will need to see how much of that improvement holds across their own projects. Faster geometry generation is useful, but the savings need to survive correction and review.

Why survey geometry demands specialized models

Mach9 says it has developed and trained its own AI models to achieve the centimeter-level accuracy required for construction and infrastructure work. Baikovitz describes development since 2024 as focused on converting imagery and combined image-and-laser-scan datasets into survey-grade geometry, including curbs and pavement striping.

The model has to locate those features and represent them geometrically. A curb boundary can look convincing on screen and still be unsuitable for a survey if its position is wrong.

Images and laser scans supply different evidence. Images reveal appearance, including painted markings; laser measurements provide spatial information. Combining them offers a way to connect visual interpretation with measured geometry, although TechCrunch’s report doesn’t explain how Mach9 does that.

The report also leaves out model architectures, training-set composition, dataset size, inference requirements, and accuracy measurement methods. It gives technical evaluators little basis for judging the implementation or its performance under difficult capture conditions.

Training specialized models fits the task. Survey extraction requires geometry with domain-specific meaning and constraints. General-purpose visual recognition alone doesn’t establish that a system can produce work a licensed professional will accept.

For AI engineers, Mach9 offers a concrete example of a narrowly defined objective: converting captured site data into editable engineering geometry. Correctness and usability determine the product’s value.

The review tools are part of the product

Mach9’s system creates a first draft that engineers refine. Baikovitz says users can draw, remove geometry, adjust vertices, and assess the quality of the AI-generated model.

Those controls matter because someone has to take responsibility for the deliverable. Baikovitz explicitly acknowledges the industry’s liability and the need for licensed professionals who understand what they’re reviewing. Mach9’s described workflow leaves that judgment with the engineer.

Fast generation can still leave expensive cleanup. Editing tools can make imperfect predictions worth using if they reduce the time needed to correct them. A vertex-adjustment tool, for example, lets an engineer repair a boundary locally rather than redraw an entire feature. Removing an incorrect extraction is just as important.

The practical measure is how quickly a reviewer can find and fix errors while keeping the correct work. Teams building similar systems should evaluate the model and interface together: prediction quality sets the starting point, and the review tools affect how much effort remains.

Mach9’s reported usage suggests customers are willing to put the software into project workflows. Nearly 100 private-beta customers and more than 2,500 projects are meaningful adoption signals. They don’t establish revenue, retention, or how much manual correction each project required.

Risks and limitations

The biggest gap is performance evidence. The reported 10-times speed improvement is a company claim. TechCrunch doesn’t provide a benchmark protocol, a task-by-task breakdown, or an independent comparison.

Buyers should establish what the comparison covers: initial extraction, drafting plus corrections, or the full path to an approved survey. Each measures a different part of the cost.

The accuracy claim needs a tighter definition too. “Centimeter-accurate” doesn’t specify average positional error, a percentile, or maximum deviation. The report doesn’t explain which feature types or capture conditions support that result. Buyers need acceptance criteria matched to their deliverables.

Professional review remains necessary in the workflow Mach9 describes. Engineers refine the draft before putting their names behind the work. That limits how much labor can be removed and makes review effort central to any return-on-investment calculation.

Integration is another unanswered question. The report doesn’t identify supported exchange formats, coordinate-system handling, APIs, or interoperability with existing Autodesk and Bentley workflows. For firms with established CAD environments, those details can determine whether faster extraction saves time or shifts work into conversion and cleanup.

Data governance is also unspecified. The source provides no details about deployment options, retention policies, or whether customer data is used for training. Technical decision-makers will need answers before submitting project datasets.

These omissions aren’t evidence of poor performance. They leave buyers with due diligence that a funding announcement can’t resolve.

What the funding needs to deliver

Baikovitz says Mach9 will spend the money on product improvements, hiring industry experts, and expanding the organization to meet demand. Industry hiring is especially relevant here. Survey-grade software needs people who can define acceptable geometry, recognize consequential errors, and understand engineering review.

Mach9 is entering a construction software market attracting investment across several workflows. CivilGrid raised $26 million in August 2026 for its underground infrastructure mapping business. Blueprints AI was featured by TechCrunch on October 12, 2026 with claims of generating permit-ready architectural plans.

Mach9’s narrower focus gives prospective customers a straightforward test: use representative captured data, generate a draft, track every correction, and compare total effort with the existing workflow. Check accuracy against the project’s requirements.

If those results support Mach9’s claims, the case for adoption is strong. Surveying firms already have data to process and deliverables to produce. They’ll buy based on how much drafting and correction the software can reliably remove.