MeltPlan / Berkeley, CA / Est. 2025
Company Profile — Vol. 01
MeltPlan
The AI planning engine quietly moving construction's loudest decisions back to where they belong - before anyone pours concrete.
The AI planning engine quietly moving construction's loudest decisions back to where they belong - before anyone pours concrete.
It is a Tuesday morning in Berkeley and somewhere on a contractor's monitor, a building code clause is being pulled apart line by line by a machine that knows exactly which page it came from. The clause is small. The stakes are not. A misread here ripples six months downstream as a change order, a redesign, a furious email at 11pm. MeltPlan is the company trying to make sure that email is never written.
The company sells what it calls a "planning engine" for construction - software for the phase before the loud one. Preconstruction is the meeting-heavy, spreadsheet-littered period where architects, contractors, code consultants, and owners argue about what is buildable, what it costs, and what the city will allow. MeltPlan's pitch is that this is the highest-leverage moment in the entire project lifecycle, and the one most under-served by software.
"Construction fails when preconstruction teams fragment and commit early with incomplete information." — MeltPlan founding thesis
Kanav Hasija, the CEO, has done this before. He co-founded Innovaccer, a health-data platform that grew into a $3B company by stitching together fractured medical records into something legible. Construction, he likes to say, has the same problem with different costumes - 95% of its data lives in documents and images, and almost none of it talks to itself.
Tanmaya Kala, the COO, came at the problem from the other side. She was a Project Executive at DPR Construction, one of the largest builders in the United States, where she watched first-hand how a thousand small preconstruction decisions calcify into one expensive surprise. The pair met around a question that doubles as a company mission: what if construction's most chaotic phase had a brain?
Co-founded Innovaccer ($3B). Now applying the same data-unification playbook to construction's document mountain.
Former Project Executive at DPR Construction. Knows what a real RFI looks like at 6:45am.
Wrote the lead check on the $10M seed via partner Pankaj Mitra, who calls construction "the last great document-and-image industry."
MeltPlan is not selling a single chatbot in a trench coat. The product is a suite, each piece aimed at one of preconstruction's grinding tasks. Together they form a feedback loop that catches problems before they get expensive.
AI building-code research that returns project-specific compliance pathways with the source clause attached. For architects, engineers, code consultants and inspectors who need to show their work.
AI-vision quantity takeoffs paired with human experts. The model proposes; a verifier confirms; the bid scope gets sharper.
Compare subcontractor proposals like-for-like. Find the line items that don't match. Surface the apples-to-oranges before signing.
For owners and developers: impact analysis and value engineering across design options, so the cheapest yes isn't the most expensive no.
"AI in construction can't be a black box. You need traceability and guardrails." — MeltPlan product principle
The throughline is citation. Every output points back at a source - a code clause, a drawing region, a bid line. In an industry where a contractor's job can hinge on what page of which code governs a particular wall assembly, "trust me, I'm an LLM" is not a feature. MeltPlan's bet is that "here is exactly where I got that" wins.
// Illustrative. Decisions made early compound late. MeltPlan attacks the leftmost bar.
MeltPlan's pilots are not 30-person startups dabbling. The first names on the customer list are DPR Construction - one of the largest general contractors in the United States - and Innovo Group, a major UAE-based contractor. That is unusually heavy company for a seed-stage product, and it tells you something about both the product's maturity and the founders' rolodex.
For enterprise contractors, the sales pitch is durable: cut rework, cut change orders, win more bids by scoping them right the first time. For owners and developers, MeltPlan is a way to argue with a contractor's numbers using something other than instinct.
The company's stated goal is to make construction boring. That is the marketing line and also the engineering choice. Boring means predictable. Predictable means budgets that hold and timelines that land. The opposite of boring construction is the construction you've actually been on the receiving end of.
The product is bilingual by design. The AI talks fluently to architects (drawings, BIM models, code clauses) and fluently to contractors (takeoff quantities, bid line items, schedule logic). Most tools in the space pick a side. MeltPlan refuses to.
And the citations matter. In construction, blame travels fast and far. A tool that prints an answer without a source is a tool no one signs off on. MeltPlan's modules sit on a substrate that always knows which clause, which drawing, which subcontractor line produced which output. That alone changes the conversation in a preconstruction meeting.
"More than 95% of data in construction is documents and images." — Pankaj Mitra, Bessemer Venture Partners
Return to the Berkeley monitor. The code clause is still being pulled apart, but now the room around it knows three things it didn't know an hour ago: which assembly is non-compliant, which subcontractor's bid quietly omitted that detail, and which design alternative the owner can live with. The change order that would have been written at 11pm six months from now will not be written, because the conversation that produces it is happening today.
That is, in the end, the whole MeltPlan idea. The drama of construction does not actually live on the jobsite. It lives in the weeks before, in the documents nobody had time to compare and the clauses nobody had the bandwidth to read. MeltPlan moves the drama forward, into a room with coffee and screens, where it is cheaper, quieter and easier to fix.
Boring, on purpose.