×
Primepoint
AI knowledge graph that reads and understands construction drawings end to end
Last verified August 17, 2026 · Updated daily
What Primepoint is building
Construction projects run on drawings. A large commercial build can have hundreds of technical drawings, each cross-referenced to schedules, specifications, RFIs, and submittals across dozens of documents. A project manager typically spends enormous time manually reconciling these: checking that a tag on a mechanical drawing matches its specification, that a revision on drawing A19 has been reflected in drawings B04 and C11, that the scope in the RFI actually aligns with what is shown. Primepoint automates this. Its AI knowledge graph connects every element in every drawing to every relevant document, then surfaces discrepancies, version conflicts, and unanswered questions that would otherwise stay buried until they cause delays or cost overruns. The AI assistant 'Marvin' answers project questions in natural language, grounded in actual project documents rather than a general model. It integrates with Procore, Autodesk Construction Cloud, and other standard GC tooling. Customer data is never used to train external models.
Why this matters
Construction is a $13 trillion global industry running on PDFs. The average large commercial project has hundreds of drawing revisions. The average project manager spends hours per week doing reconciliation that produces no new value. The average dispute or delay can be traced to a missed discrepancy between a drawing and a spec that nobody caught in time. AI has struggled with this problem because it requires more than language understanding: it requires reading linework, interpreting spatial relationships, following cross-references, and connecting that information to structured project data. Foundation models cannot do this reliably at project scale. Primepoint has built domain-specific computer vision and knowledge graph infrastructure to handle it. The fact that Yann LeCun, one of the three people most responsible for the modern AI era, made a personal angel investment is not just a credential. He was a colleague of co-founder Lubomir Bourdev at Facebook AI Research. He put his own money in because he understood the technical approach from the inside.
Open roles at Primepoint
3 positions we're tracking. Roles are re-checked daily and removed when filled.
Know when Primepoint is hiring before anyone else
A role stays uncontested for about four days. Here's the window — and where we put you in it.
From $9/month, cancel any time.
Watching Primepoint
0 applicantsRole spotted & verified
1You get the alert
1You've applied
~8Hits the job boards
250+Hiring outlook
High. Small team of 8 with $10M raised and an explicit customer base of large commercial general contractors. Engineering and commercial expansion are the stated use of capital.
Working at Primepoint
Primepoint: AI knowledge graph that reads and understands construction drawings end to end. Founded , currently <50 employees. At this stage, expect broad remit, direct access to founders, and equity that still means something if the company works out.
Most Primepoint jobs are based in San Mateo.
How to actually get hired at Primepoint
Why applying the normal way doesn't work
With only <50 employees, Primepoint doesn't have dedicated recruiters. Founders review applications between running the company. The challenge isn't competition — it's visibility. Direct outreach wins.
Who to contact at Primepoint
Founding team
Bourdev is a founding member of Facebook AI Research and one of the architects of modern computer vision. He built the original object recognition system that runs on every photo and video uploaded to Facebook and Instagram, a system operating at a scale most engineers never touch. Before Primepoint, he co-founded WaveOne, a deep learning video compression company, which Apple acquired in 2023. He holds more than 100 patents and has over 100,000 citations on his published research. His PhD is from UC Berkeley in Machine Learning. He is from Plovdiv, Bulgaria, studied at Brown University as an undergraduate, and has spent his career at the intersection of research-grade AI and real-world deployment. Bourdev is one of the most cited researchers in computer vision. His early work on human pose estimation (the "poselets" paper at Berkeley) is foundational to modern visual understanding systems. He has published across CVPR, ICCV, and ECCV. His personal website at lubomir.org is the best entry point for his research arc. He is not a blog writer or a pundit. His ideas live in papers and in products. Yann LeCun investing personally is the peer review equivalent: one of three people most responsible for modern deep learning chose to back him with personal capital. Bourdev grew up in Bulgaria during the post-Soviet transition, an environment where resourcefulness and technical self-sufficiency mattered early. He moved to the US for Brown University, then stayed for Berkeley and the FAIR founding. The path from Bulgarian childhood to founding Facebook's AI research organization to building construction intelligence software is not an obvious line, but it makes sense when you see the thread: he has always worked on problems that require understanding the visual world at scale, from human bodies to video compression to technical drawings. WaveOne's Apple acquisition in 2023 means he has already been through one full cycle from founding to exit.
Palo joined Trello as employee number five. He spent nine years at Trello and then Atlassian after the $425M acquisition, eventually becoming Director of Product and leading Trello's transition to enterprise. After a sabbatical in 2022, he joined Uber to take grocery delivery from zero to one. The career arc is a masterclass in product scale: he has been inside a product at single-digit headcount and inside a product at tens of millions of users, and he knows what changes at each stage. He has been building AI-native products in construction since before Primepoint was announced, which is why Bourdev recruited him specifically. He gave a detailed podcast interview on the ScaleUp & Up podcast (July 2024) about product development at Trello, Atlassian, and Uber, and his thinking on construction tech. That episode is worth listening to before outreach. He discusses letting go of ownership in a fast-scaling product, the emotional toll of acquisition, and his specific view on what goes wrong when startups try to hire the conventional "middle" rather than the extremes. This maps directly to Eigen's "U-shaped" team thesis, which suggests it is a shared philosophy in the Bay Area founding community right now. Joining Trello as employee number five is one of the more remarkable early bets in modern SaaS history. Trello had $0 in revenue when Palo joined and was acquired for $425M within a few years. He rode that arc all the way through enterprise product management before walking away for a sabbatical. The sabbatical itself is notable: a 2022 pause, taken deliberately, before moving to Uber, then founding Primepoint.
What to show them
This is Bourdev's domain. The product is built on specialized computer vision and knowledge graph infrastructure for technical drawings. As the platform expands to new project types and document formats, the ML team needs to grow. Bourdev spent his career building systems that understand visual content at scale: from Facebook's image recognition to WaveOne's video compression. He will hire people who think about visual representation and spatial understanding at a research level, not just application level. Core skills: Computer vision, knowledge graphs, document AI, PyTorch or JAX, spatial reasoning, Python, experience with structured document understanding or OCR-adjacent problems. Proof of work: Take any publicly available set of construction drawings (many architecture schools post these) and write a short technical memo describing how you would approach connecting drawing tags to a specification document using ML. Include the specific failure modes you'd expect and how you'd test against them. No code required; the thinking is the proof.
A cold email that works at Primepoint
What Primepoint screens for
The press release explicitly states the capital will fund platform development and expansion of the customer base to large commercial general contractors across the US. Sectors named for expansion: data centers, residential developments, and education facilities. That is a product roadmap and a sales roadmap announced simultaneously. Navitas Capital is the lead investor and a construction technology specialist. Their portfolio is concentrated in proptech and construction software. They do not invest in construction-adjacent SaaS; they invest in construction-native tools. That focus means they push portfolio companies toward deep GC and subcontractor integrations, enterprise sales motions, and construction-specific compliance and data standards. That is a hiring thesis: enterprise AEC sales, solutions engineering, and construction domain expertise. NextView Ventures focuses on seed-stage companies solving everyday problems through software. Their involvement alongside Navitas suggests Primepoint is being positioned as infrastructure software with a wide deployment surface, not just a point solution for a narrow niche. That shapes the growth and product roles needed. Yann LeCun's check is a technical credibility signal in a market where credibility is hard to establish. Large GC procurement teams are conservative. Having the name of the founder of modern deep learning attached to the round makes enterprise sales conversations easier and opens doors to technical partnerships. One live role posted via Dover: app.dover.com/apply/primepoint.
Don't send a generic CV to Primepoint. Mirror the job posting's language to get past automated screening.
Don't make these mistakes
Calling this a supply chain startup (it is not). Vague enthusiasm about 'AI transforming construction.' Every contractor sales rep says that. Be specific about drawings, specifications, and the reconciliation problem.
Mistakes that kill Primepoint applications
At <50 people, a copy-paste CV is immediately obvious. It's an instant no.
Lead with their problem, not your ambition. Primepoint is focused on Construction is a $13 trillion global industry running on PDFs — show you understand that.
Don't apply and wait. The median AI application gets no response ever. One follow-up at day five roughly doubles reply rates.
Applying to Primepoint? Get the contact, not the form.
The Primepoint interview process
3 stages · 7 days typical · take-home: yes · modelled from similar companies
We don't yet have verified candidate reports for Primepoint. What follows is the typical process for a <50-person AI company — treat it as a model, not confirmed detail.
Interview stages
Intro Call
Video call · 30 min
Culture fit and role expectations
Founder or hiring manager
Technical Deep Dive
Video call or in-person · 60 min
Past projects and problem-solving approach
Technical founder or lead
Final Round
In-person or video · 45 min
Team fit and offer discussion
Founding team
Primepoint take-home assignment
Primepoint includes a take-home exercise in their interview process. For AI roles, this typically involves a practical problem that takes 2-4 hours. Focus on clean, working code over premature optimization. They're evaluating how you think and communicate, not just the solution.
Primepoint interview timeline
At days, Primepoint's process is faster than typical for AI (10 days at this size).
Interviewed at Primepoint?
Tell us how it went — stages, questions, timeline. Takes 90 seconds and it's how this page stays accurate for the next person.
Submit your Primepoint interview experience →Primepoint jobs, frequently asked questions
How many jobs does Primepoint have open?
Primepoint currently has 3 open roles, last verified April 2026.
Does Primepoint hire remotely?
All current Primepoint roles are based in San Mateo.
What roles is Primepoint hiring for?
Primepoint is hiring across Engineering, Product. The most recent opening is Senior ML Engineer / Computer Vision.
How do I apply for a job at Primepoint?
Apply directly through the links above, or read our guide on how to actually get hired at Primepoint.
Does Primepoint respond to cold emails?
We're still collecting cold email data for Primepoint.
Who is the hiring manager at Primepoint?
At this size, hiring is usually run by a founder or department head.
How competitive is it to get hired at Primepoint?
<50-person AI companies see ~50-100 applicants per role in two weeks. The 72-hour window is your advantage.
How many rounds is the Primepoint interview?
3 stages: Intro Call, Technical Deep Dive, Final Round.
Is the Primepoint interview hard?
It concentrates on technical depth and system design rather than abstract puzzles. The stage candidates find hardest is Technical Interview.
Does Primepoint give a take-home task?
Yes, Primepoint includes a take-home assignment.
How long does Primepoint take to get back to you?
Around 7 days across the full process.
What should I prepare for the Primepoint interview?
Prepare for technical depth and system design. A <50-person startup wants proof you can ship, not that you can whiteboard.
Where is Primepoint based?
Primepoint is headquartered in San Mateo, US.
Get Primepoint roles before they're posted
A role stays uncontested for about four days. Here's the window — and where we put you in it.
From $9/month, cancel any time.
Watching Primepoint
0 applicantsRole spotted & verified
1You get the alert
1You've applied
~8Hits the job boards
250+