Picking Up the Pace: The Acceleration of Kindred AI

An interview with Kindred AI CEO, Marin Tchakarov

Kindred Sort

My first introduction to Kindred AI occurred at the Manhattan Associates booth at Promat 2019. There I learned about the company’s partnerships with automation vendors and the forthcoming Manhattan Automation Network certification. As I stated back in 2019, I found the partnership with Kindred and the integration with Kindred Sort particularly interesting. At the time I became aware of the deployment of Kindred AI at Gap Inc. Fast forward to the Spring of 2020, and I learned about Gap Inc.’s plan to triple its number of Kindred item-picking robots and then earlier this summer about American Eagle’s expansion of its Kindred SORT robot fleet. I had the opportunity to speak with Kindred AI’s CEO, Marin Tchakarov shortly thereafter, and felt it was time to learn first-hand about Kindred AI and what appears to be a rapidly growing installed base of solutions.

Clint:  Marin, congratulations on your recent appointment as CEO of Kindred. What was your career path that led you to your role at Kindred?

Marin:  I have been in high-tech for decades, advancing to leadership roles in the early 2000s. Prior to Kindred, I was in the IoT space, serving as CFO at Jawbone and Pebble. Most recently, I served as Chief Operating Officer here at Kindred during a period of robust growth.

Clint:  Indeed, Kindred’s recent expansions at Gap, Inc. and American Eagle suggest that Kindred itself is in a period of robust growth. The ongoing growth of e-commerce and the subsequent increase in item picking certainly presents an opportunity for intelligent automation. But it appears to be a complex process to automate. What is technologically different about item picking? And what technologies enable it to be operationally feasible?

Marin:  The unstructured, chaotic and dynamic environment of item picking makes it a complex problem to solve, especially with the accuracy and speed required in a production setting. The Kindred AI powered platform integrates machine vision, grasping and placement technology that autonomously solves complex real-world commercial problems in the e-commerce fulfilment center.  Kindred Sort focuses on piece picking within a sortation solution and it is an embodiment of Kindred’s patented platform.

Clint:  What capabilities do you believe differentiate your solution from others that I have seen on display at trade shows and elsewhere?

Marin:  There are a few characteristics I can mention that differentiate Kindred in the market. The company was founded in 2014 and we have been developing this platform for 6 years. This has generated a ton of IP (intellectual property) provided capabilities, including AI powered machine vision with the ability to singulate separate items in a confined space, infer shapes and sizes and where certain attributes, such as bar codes, reside on an item. There is an extensive number of things that go into grasping and placing technology. For example, how to calculate best path for grasping; how to avoid falsely executed commands such as zero picks (failed attempt) or errored attempts (picking two items instead of just one). The algorithms require a lot of fine tuning to minimize errors on “both sides of the fence.” Performance must occur at a level of accuracy necessary in production setting. Uptime must also be extremely high. The vectors that allow us to operate consistently are exceptionally high accuracy and uptime together with the volumes of speed and throughput. And somewhat ironically, as you asked about Kinred’s differentiating characteristics, the real answer is the people on the team behind the solution. We really have an incredible team. It’s our people behind the robots.

Clint:  What types of fulfillment profiles do you think can best benefit from your picking solution, and why? Where is it not as applicable?

Marin:  Our current product is ideally suited to retailers with a significant presence and business in e-commerce because of the uniformity of the inventory and packaging in a typical e-commerce fulfillment operation – more specifically, merchandise in polybags with bar codes in a fairly uniform way. Inversely, what provides difficulties for us is to have items that come in a variety of different ways that do not conform to certain requirements. For example, basketballs, loose clothing, and suits or dresses on hangers. To be clear, the Kindred system can handle these items, but we don’t want to potentially damage the robot or merchandise. It’s also worth noting that our ability to handle deformables such as apparel is a massive differentiator from the rest of the competition.

Clint:  To what degree has your solution been deployed at customer sites thus far? What are you plans for 2021?

Marin:  I can say that we will have roughly tripled our fleet from last year to the end of this year. This year we will be well into the hundreds and we expect to continue along this growth vector in our future. We are barely getting started. There is an ocean of white canvas in front of us.

Clint:  Thanks Marin. Please keep Logistics Viewpoints posted on future developments at Kindred.

Leave a Reply

Your email address will not be published. Required fields are marked *