In the rapidly evolving landscape of warehouse management, Gather AI stands out as a pioneering force, leveraging the power of artificial intelligence and autonomous drones to transform inventory monitoring. We had the opportunity to sit down with Sankalp Arora, Co-Founder of Gather AI, to delve into the journey from a groundbreaking concept to a fully realized solution. From the eureka moment at Carnegie Mellon University to securing a significant $17M funding round, Sankalp shares the pivotal moments and innovative strides that have positioned Gather AI as a leader in the field. He also offers insights into the unique capabilities of their technology, future plans for scaling, and invaluable advice for aspiring entrepreneurs in the AI and robotics domain.
Sankalp, can you walk us through the journey from concept to execution with Gather AI? What were the pivotal moments that led you to focus on AI for warehouse inventory monitoring?
Our digital world and all the SaaS products we have today work on structured data. Large language models (LLMs) enable us to make unstructured data useful, however, there is an opportunity to tap a large pool of data that is not digitized, which I call physical data. I wanted to build to solve the problem of generating insights on physical data. The “eureka” moment came about while working toward my PhD at Carnegie Mellon University, developing the world’s first guaranteed safe full-scale autonomous helicopter with my future co-founders, Daniel Maturana and Geetesh Dubey.
I was standing on FBI training grounds in Quantico, where I watched our full-scale autonomous helicopter come in and land. That helicopter had just covered 10 kilometers of land in under three minutes and built a 3D map of the environment. That led me to realize that robots are powerful large-scale data-gathering machines, and can be leveraged to digitize the physical world. Our project won the Howard Hughes award, AUVSI Xcellence award, and was nominated for the Collier Trophy. The Department of Defense funded a customer discovery process for the application of our tech. Through over 175 customer discovery interviews and a partnership with dnata, we were able to see an urgent and compelling problem in inventory monitoring, which led to the founding of Gather AI in 2017.
With the recent $17M funding round, how do you plan to scale Gather AI’s technology? Are there specific areas of the warehouse operations you’re targeting for further innovation?
We’ll use this funding to scale operations as we continue to grow rapidly by solving supply chain issues with richer data and AI.
In terms of specific innovation areas, we’re focused on AI-enabled vision capabilities. Our computer vision engine is a core tool for warehouse operators to understand the state of their inventory, for example, how many items are in a warehouse, whether they’re damaged, whether they are stacked right, etc. Our AI software brings us to the forefront, and with our solution, warehouses can decrease their inventory errors by 66% on average. Barcodes disrupted the 80s and 90s supply chain space, and computer vision is disrupting it now.
We are investing in bringing the richest image-to-inventory data to our customers across multiple warehouse sites. We recently launched industry-first inferred case counting and location occupancy capabilities which enable warehouses to get automated, digitized counts and location utilization reports, unlocking higher on-time shipment rates while reducing dedicated counting labor. You will see more of such features coming from Gather AI.
Today, we use drones to gather image data, which our AI analyzes. Our roadmap is built to enable us to use other devices to collect the images and generate insights. We also want to bring this visibility to areas within the warehouse—on the ground, on loading docks, and more.
Gather AI is described as a leader in computer vision-based AI. Can you elaborate on how your technology differs from other solutions in the market, particularly in terms of accuracy and efficiency?
We differ in three major ways:
- We make cobots (collaborative robots), making the current workforce in warehouses into superhumans. Efficiency/speed is the keyword here, enabling a single person to do inventory checks on 900 pallets/hour, where they only used to be able to do 60 pallets on average.
- Our system provides a rich set of inventory insights like case counts, occupancy reports, empty detection, label reads and barcode reads, while most of the industry is focused on just providing a better barcode reader. We also read barcodes, but can read all in a location in a single image leading to 4-5x faster barcode reading alone, while most in the industry read one barcode at a time.
- Needing no infrastructure changes or additions, we’ve developed the solution to suit existing warehouse environments. Our AI algorithms ‘fly’ the drone autonomously in the warehouse with no WiFi, infrastructure, or label changes needed. The AI algorithm also analyzes text and barcodes on labels, counts boxes, and estimates occupancy. Of note, our solution can read 3x smaller barcodes than most standard engines. The algorithm improves as more and more warehouses are scanned.
Drone-powered inventory systems are a significant innovation in supply chains. Could you explain how they work in a typical warehouse environment and what makes them more effective compared to traditional methods?
With our warehouse inventory monitoring solution, warehouse employees no longer spend long, tedious hours doing manual inventory with forklifts, and there’s less likelihood of misplacing products (no overordering, delayed shipments, or “fire drills” looking for lost inventory). The warehouse manager can view inventory data in real time from a web dashboard and easily identify and fix inventory exceptions, even creating a to-do list for their teams.
With our current drones, customers can do barcode scans, verify quantities, and visually verify the state of the product 15x faster than manual methods. We’ve helped facilities go from 90-day case counts to just 2.5 days, collecting rich data autonomously. Our customers have drastically decreased inventory loss and shrinkage because our drones can scan warehouses more quickly, so they know where everything is in the warehouse.
Our solution is currently deployed in warehouses across third-party logistics, retail distribution, manufacturing, food and beverage, and air cargo, and it can be applied to any warehouse with racking.
Looking ahead, how do you see AI and automation evolving in the business landscape over the next five years, and what role will Gather AI play in this evolution?
Generative AI will make prediction and analytics on warehouse data more accessible. It will enable data insights to be available on-demand through natural language interfaces and help us make executive decisions in real time.
However, the reliance on that data means it needs to be accurate, which is where we come in. We enable supply chain operators to know what’s on the floor in real time and make the source of data traceable. Operators will be able to see an image of a package, its exact location, and its condition, vs. just seeing a status email. Gather AI makes that enhanced visibility as easy as the press of a button and powers the next generation of optimizations in the supply chain space.
What are some of the biggest challenges you’ve faced while integrating AI technologies into traditional warehouse operations, and how have you overcome them?
Today warehouses are unstructured. There are lighting problems, labels and boxes come in all shapes and sizes, there’s poor network infrastructure and more which can cause visibility challenges. We have overcome this by collecting warehouse data to make a moat and developed the product for five years in warehouses. Our in-warehouse, data-intensive development approach has led us to a product that needs no infrastructure changes in warehouses while having the ability to provide best-in-class data insights.
Lastly, as a leader and innovator in a rapidly advancing field, what advice would you give to young entrepreneurs aspiring to venture into AI and robotics?
At Carnegie Mellon’s Field Robotics Center, we had this adage, “Don’t focus on the tech. Focus on the problem you’re solving.” The problems AI and robotics can solve have broadened, specifically with transformer networks powering large language and diffusion models coming forward in the last few years. While technology is a powerful enabler to solve things that people accept as hard facts of life, be sure to focus on the problem you’re solving, and ensure there’s an appetite to address that hard fact of life your AI is solving. You will make magic happen.
