Legora for physical engineering firms
HQ: San Francisco, CA, United States | Founded: 2026 | Employees: 6 | Stage: Seed (Y Combinator S26) | Website: https://www.usetorus.com
Torus builds AI engineering tools for the firms that design capital projects, and describes itself as 'Legora for physical engineering firms'. Its customers — engineering firms and project owners — use Torus to generate complete engineering packages, including complex diagrams across several disciplines, with the aim of sharply cutting engineering cost and time. The company's framing is that infrastructure projects spend months or years on engineering alone, and that every week of schedule compression on a billion-dollar project is worth millions. Torus reports working with multiple Fortune 500 companies, major engineering contractors and infrastructure owners across the United States on critical infrastructure including data centres and energy installations. The team works in person in San Francisco and travels to customer sites in Kansas City, Tulsa and Houston. Founded in 2026 by Marcus Lima (chief executive) and Rahul Thayil (chief technology officer), friends for 13 years who lived together for five. Lima previously founded Heimdal (YC S21), a direct air carbon capture company that raised more than $12M, including from Sam Altman and Marc Benioff, and built the largest carbon capture plant of its kind, giving him years of direct experience working with engineering contractors. Thayil previously founded Orion, an AI threat intelligence company that raised more than $4M, and has spent more than 10 years building real-time anomaly detection over messy physical data.
Torus applies the 'AI for professionals' playbook to physical engineering, which is a smarter target than it first looks. Infrastructure projects spend months or years on engineering before anything gets built, and every week saved on a billion-dollar project is worth millions, so the return on the software is unusually easy to argue for. The 'Legora for X' framing signals the template: pick a document- and drawing-heavy professional workflow, wrap AI agents around it, and charge against the labour it replaces. The hard parts are that the buyers are cautious engineering and utility firms, established vendors own the workflow, and the accuracy bar is absolute — a diagram with a subtle error in a power plant design is not a small bug.
Browse: Home | Blog | About | All Startups A-Z | View full profile on StartupWiki