Six years and $1.5M of their own money to build an autonomous floor robot

Executive overview

Most robot vacuums fail at basic intelligence: they can't map spaces, distinguish surfaces, or remember where they've been. Navneet Dalal and Mehul, co-founders of Matic, spent six years solving this from first principles — building a camera-first, software-heavy autonomous floor robot.

Their prior exit (Flutter, acquired by Google) taught them two hard lessons: cool technology without a real problem is just a feature, and software-only solutions are arbitrarily limited. Matic is the deliberate correction.

The insight: a truly autonomous indoor robot must behave like a human — patrolling, adapting to surfaces, and remembering what it has already cleaned.

The problem with existing robots

• Today's robots have no persistent memory — restarted, they re-clean the same area
• Existing products handle suction well but vision poorly; one destroyed a $2,000 rug
• The floor-cleaning category grows 25% year-over-year, yet consumers tolerate inferior products
• No robot could reliably distinguish a rug from a hard floor, let alone adapt its cleaning method

The Matic approach

• Level-five autonomy benchmark: the robot should clean like a human, not just follow a pattern
• Camera-only sensor strategy — adding sensors multiplies software complexity, calibration needs, and failure points exponentially
• Two RGB cameras plus algorithms mirror how humans navigate; fewer sensors means a more scalable platform
• Robot should patrol continuously, detect dirty spots, and switch between vacuuming and mopping as the surface demands

Building in hardware

• First prototype: a Makita vacuum motor fitted with a stereo camera, streaming images over Wi-Fi to a laptop
• 200+ prototypes built and destroyed before reaching a production design
• Bootstrapped with $1.5M of their own money; 3D printing becoming cheap was a critical enabler
• Raised $30M total; investors include Patrick and John Collison, Jack Dorsey, and Naval Ravikant

Lessons from Flutter and Google

• Flutter (gesture control for laptops) reached #1 in 73 countries on the Mac App Store with 77M gestures performed
• Paul Graham's repeated challenge — "you're technology looking for a problem" — proved correct
• Being camera-dependent inside a MacBook meant no control over frame rate, exposure, or focus — a half-solution
• Post-acquisition year at Nest was deliberate: learn hardware properly before starting Matic

Principles for hardware founders

• Solve your own problem first — the Golden Retriever rug incident was the founding insight
• Identify the riskiest assumption early (usually market risk, not technology risk) and de-risk it first
• Ask "why" three times in a row before accepting how something is designed
• Only do hardware if the problem genuinely cannot be solved in software
• Need more patience and a clear revenue model from day one