Ambi Robotics Sorts 100 Million Parcels on a Bet About Grasping

The Berkeley spinout, backed by Tiger Global and Andreessen Horowitz, is using simulation-trained AI to automate the warehouse's most stubborn task.

About Ambi Robotics

Published

The robots in Ambi Robotics' Berkeley lab have sorted more than 100 million parcels. The company's CEO, Jim Liefer, says that figure is a testament to the system's reliability. For the logistics operators buying these machines, it's a calculation about labor and throughput. In a Pitney Bowes facility in Stockton, California, an AmbiSort system nearly doubled parcel throughput during the 2021 holiday peak. That kind of performance is why the company's newer AmbiStack palletizing robots have their entire 2025 production inventory already reserved [8, 9, 11]. Ambi is betting that its simulation-to-reality AI, developed from a decade of academic research, can finally handle the messy, variable world of e-commerce fulfillment.

The academic wedge

Ambi Robotics is a hardware-and-software play built on a software-first insight. The company's core technology, AmbiOS, uses synthetic data and simulation to train robotic arms to grasp and manipulate items they have never seen before [Perplexity Sonar Pro Brief, Unknown]. This approach, pioneered in UC Berkeley's AUTOLAB under Professor Ken Goldberg, addresses a fundamental bottleneck in warehouse automation.

Founder / Key Leader Role Background
Ken Goldberg Co-founder, Advisor Professor of IEOR at UC Berkeley, head of AUTOLAB, roboticist for three decades [3, 7, 10]
Jeffrey Mahler Co-founder, CTO PhD from UC Berkeley, co-developer of Dex-Net grasping research [5, Perplexity Sonar Pro Brief, Unknown]
Jim Liefer Chief Executive Officer Over 30 years in supply chain and e-commerce operations from Fortune 50 companies to startups [4, 5, 7]

A system, not just a robot

Ambi sells integrated solutions, not standalone robotic arms. The product suite is built for specific, high-volume warehouse workflows. The AmbiSort systems, in A-Series and B-Series configurations, are designed for parcel induction and sortation into sacks, gaylords, or carts [Perplexity Sonar Pro Brief, Unknown]. AmbiStack automates the palletizing and destacking of mixed cases and parcels [7, 9, 12]. A newer offering, AmbiKit, targets kitting operations [Ambi Robotics, Unknown]. The company positions these as turnkey systems, combining proprietary hardware, the AmbiOS software brain, conveyors, and ongoing support under a commercial service model [1, 6, 12].

The capital behind the gripper

To build and deploy physical robots at scale requires significant capital. Ambi has raised approximately $64 million in total disclosed funding, anchored by a $26 million Series A led by Tiger Global in 2021 and a $32 million Series B in 2022 [3, StartupIntros, Unknown].

Round Amount
2021 Series A $26M
2022 Series B $32M

Where the wheels could come off

The warehouse automation space is crowded and capital-intensive. Ambi is not the only company trying to solve the piece-picking problem. Well-funded rivals like Berkshire Grey, Covariant, and Plus One Robotics are pursuing similar customers with different technical approaches [Perplexity Sonar Pro Brief, Unknown]. The competitive pressure is not just on technology, but on deployment speed, total cost of ownership, and the ability to integrate into complex, legacy warehouse environments.

The next twelve months

With its 2025 AmbiStack inventory already spoken for, the immediate challenge is execution. The company must successfully manufacture, deliver, and commission those systems while continuing to support and expand its sortation installed base. Key milestones to watch will be the public announcement of additional Fortune 500 customers beyond Pitney Bowes and OSM, and any movement toward a Series C round to fund further scaling.

Read on Startuply.vc