The autonomous back-of-house for restaurants.
Hey YC! We’re Arjun and Aakar, and we’re launching Marble (previously Truffle), the first AI-native operating system for the restaurant back-of-house.
TL;DR: We help restaurants control their two biggest costs, inventory and labor—saving them thousands of dollars per month in food waste, stockouts, and labor overage. We replace today's patchwork of disconnected point solutions, spreadsheets and clipboards with a single, easy to use system that actually does the work. We use computer vision and voice AI to count inventory 90% faster than manual counts, ML models to forecast item-level demand with 97% accuracy, and AI agents that run dozens of end-to-end workflows autonomously, including ordering, invoice reconciliation, scheduling, food prep, menu engineering, and more.
The result: We're live and scaling rapidly across multi-unit operations in the US and Canada. You can find us today in A&W Restaurants, The Kati Roll Company, Sophie’s Cuban, Kokomo NYC, HolyShred, Dead Letter No. 9 and many more. We’ve driven a 40% reduction in food waste, eliminated stockouts, and saved managers 30+ labor hours per week.
Check us out here: www.joinmarble.ai
Our ask: If you run or know someone who runs a restaurant, café, bar, or hospitality group, we'd love to talk. You can reach us at founders@joinmarble.com.
Everyone knows someone who runs a restaurant, bar, or cafe - we’re also giving away $500 (or an Oura Ring) per referral! www.joinmarble.com/refer
The Problem
Restaurant operators face an uphill battle. Ultra-competitive markets, rising food and labor costs, unforgiving hours, and after all of it, they take home 5% in net profit (if they're lucky).
A restaurant can be full all day and night, but it's the back office operations that determine whether that money hits the bottom line. Today that back office still runs on clipboards, spreadsheets, and guesswork. Inventory gets counted by hand, if at all, so no one has a real-time picture of what's actually on the shelf. Orders get placed from memory, so operators either over-buy (contributing to $150B+ in annual food waste) or under-buy and stock out on the exact items guests came for. Food prep and schedules are built on best-effort guesses, and invoices go unchecked while suppliers quietly raise prices line by line.
Every one of these is a margin decision, and every one is being made blind, leading to errors that compound every day, across every location.
The Solution
Marble runs the entire back office for restaurants. This centers around managing the two biggest cost centers (inventory and labor) and works in 4 phases:
- Capture: Computer vision + voice AI through our mobile app allow operators to count inventory 90% faster than manual counts, with 99% accuracy. No more hours spent counting items in a walk-in freezer with a clipboard (IYKYK). Invoices and deliveries are captured the same way.
- Forecast: ML models predict item-level demand for the weeks ahead at 97% accuracy. In addition to that, LLM-powered agents scan the signals a spreadsheet never sees - local events, weather, news, reservations, holidays - and adjust the forecast for what's actually happening around each location.
- Act: A set of AI agents run the workflows that actually move the needle on profitability: placing optimized orders, building demand-informed prep sheets, drafting schedules, proactively tracking and flagging waste, reconciling invoices, engineering menus, and more.
- Learn: This industry runs on tribal knowledge. We have built the product as a closed-loop system from day one: every count, correction, and order flows back into the platform as context that our agents carry forward, so Marble learns how each location actually runs, adapts to its patterns automatically, and gets sharper every day.
Team
We’ve been close friends for nearly 10 years, since our freshman year at Michigan, and we're having a blast taking on this industry.
Aakar grew up managing the back office for his family's multi-unit quick service restaurants, where he saw the problem up close. He later covered the restaurant sector as an investment banker at Goldman Sachs before leaving to lead growth at an applied-AI startup in healthcare.
Arjun is a second-time founder and was previously a Senior Software Engineer at Bloomberg, where he led AI/ML initiatives on the Electronic Trading team. At Bloomberg, he built out complex trading systems that processed tens of millions of dollars daily.
