He teaches buildings to read their own paperwork - so the field finds out about the conflict before the concrete does.
▲ Twenty years, three companies, one stubborn question: can construction documents talk to each other?
Open any active construction project and you will find a paper avalanche: submittals, requests for information, specifications, contracts, design drawings - documents that contradict each other in ways no one notices until a crew is standing in the field holding the wrong instruction. Ivan Panushev runs TwinKnowledge, a New York company whose AI agents read all of it. Not skim. Read. Then they flag the scope conflict before anyone pours.
Founded in 2023 and headquartered at 575 Lexington Avenue in Manhattan, TwinKnowledge builds what Panushev calls AI copilots for design, engineering, and construction professionals. You ask a question. The copilot answers from your organization's own body of knowledge, and it shows you the source. No hunting through PDFs at 11pm.
In April 2025 the company closed a $3.7 million seed round led by Camber Creek, the real-estate-focused venture firm, with Great Wave Ventures joining and former Autodesk Construction general manager Jim Lynch taking a board seat. The customer list reads like a cross-section of who actually builds in America: the U.S. Space Force, Toll Brothers, SHoP Architects, and Sound Transit.
The pitch is unglamorous and exactly right. Construction runs on documents. The documents are a mess. The mess costs money, time, and rework. Panushev has spent his career arguing that software can clean it up - and this is the version where the software finally can.
We are pioneering the integration of AI agents into the AECO industry - using computer vision and large language models to simplify construction document analysis.— Ivan Panushev, on TwinKnowledge's $3.7M seed
The AI agents examine millions of data points across submittals, RFIs, specs, and design documents - pulling precise, real-time answers out of paperwork that used to require a war room.
Every document is automatically validated against project requirements and construction standards, so compliance stops being a manual scavenger hunt.
The system spots scope conflicts between contracts, documents, and specifications before field work begins - the difference between a comment and a costly change order.
Panushev did not arrive at construction AI by accident. He earned a Doctor of Design from the Harvard Graduate School of Design in 2008, stacked on top of a master's from the same school and an undergraduate degree from Harvard College. He stayed close enough to academia to hold a faculty seat at the GSD, and he postdoc'd at Georgia Tech's Digital Building Laboratory, where he co-authored sections of the National BIM Standard - the rulebook for how building data is supposed to be structured.
Then he tried to build the structure himself. As co-founder and CTO of Horizontal Systems, he made cloud software for multi-discipline model collaboration. In 2011 Autodesk bought it, and the technology became the foundation of BIM 360 - which grew into Autodesk's fastest-growing cloud offering. Panushev stayed on as a senior platform manager, herding intelligent object data across more than 40 different 3D file formats.
His resume refuses to sit still. He ran strategic partnerships at MakerBot during the desktop 3D-printing boom. He co-founded Inpher, a privacy-preserving machine learning company, and sat on its board. He served as Principal Partner Solutions Architect for Engineering, Construction, and Real Estate at Amazon Web Services, the role that put him at the center of how the entire industry was moving to the cloud.
Two McGraw-Hill books on building technologies carry his name. So does a seat on the U.S. National BIM Program Steering Committee at the National Institute of Building Sciences. The through-line is unmistakable: every job was a different angle on the same problem. TwinKnowledge is the angle where the tooling finally caught up to the ambition.
Caps three Harvard degrees and turns toward how buildings store and share data.
Co-authors sections of the National BIM Standard at the Digital Building Laboratory.
The startup he co-founded and led as CTO becomes the foundation for BIM 360.
Manages model collaboration and object data exchange across 40+ 3D formats.
Leads strategic partnerships at MakerBot; co-founds privacy-preserving ML company Inpher.
Principal Partner Solutions Architect for Engineering, Construction, and Real Estate.
Sets out to make every organization's construction knowledge instantly queryable.
Great Wave Ventures joins; ex-Autodesk Construction GM Jim Lynch takes a board seat.
Our technology significantly reduces errors, accelerates project timelines, and raises the bar for accuracy, efficiency, and productivity.— Ivan Panushev
Panushev's wager is simple to state and hard to deliver: make a building's accumulated knowledge instantly answerable. Ask a question, get a precise and sourced reply, and let scope conflicts surface before they ever reach a crew in the field. It is the same problem he chased at Horizontal Systems, at Autodesk, at AWS - now with agents that can finally read the whole library.