The AI company that starts with the packing tray

A car begins as a problem of arrangement. Before its parts become a vehicle, they must be collected, packed and delivered in a form that makes assembly possible. At one manufacturer, preparing those collections meant laborious manual work, mistakes and damaged parts. The interesting thing about inmind.ai is that it found an artificial-intelligence story inside this decidedly unglamorous job.

The collections are called Carsets. Parts go into trays and shelves; their shapes and the available space impose rules. inmind.ai built Pick Pack to model configurations and find more efficient packing arrangements. In its account of the project, the company identifies Rolls Royce and reports preparation capacity increasing from two sets daily to twelve, alongside reduced part damage.

That number deserves attention precisely because it is narrow. It describes one company-reported project, one task and one measure. It offers a more useful way to understand this business than a catalogue of fashionable technologies: look for the awkward piece of work, then examine what the software changes.

First, understand the tray

A packing simulation has to respect what a part actually looks like. Space constraints and design variations matter. Pick Pack reproduces different configurations so planners can compare layouts before committing to them. The intervention happens upstream of assembly, where a seemingly small packing decision can create work for everyone downstream.

That helps explain inmind.ai’s three main service areas: AI, robotics software, and digital twins. A digital twin is a virtual representation of a physical asset or process. The company offers modeling, scenario testing and continuing enhancement using operational data. Here, the appeal is practical: try an arrangement in software before making people carry it out.

The lesson a reader can borrow is methodological. Choose a task with observable constraints and a result someone can count. Packing capacity qualifies. So does time spent finding a document. A broad ambition such as improving productivity becomes easier to judge when it has an address inside the working day.

A robot needs more than a brain

Consider idealworks, the Munich robotics business. inmind.ai describes two related assignments: software for an autonomous mobile robot, and a platform to coordinate robot fleets. The first concerns navigation, obstacle detection and operating rules. The second concerns what happens when several machines need the same space, at the same time, for different jobs.

A capable robot can still become an inconvenient colleague. Fleet orchestration allocates tasks, coordinates movement and connects with warehouse management and enterprise resource planning systems. inmind.ai’s case describes real-time monitoring and efforts to reduce idle time. The value lies partly in the connections between machines and the business systems directing their work.

This engineering reaches into simulation, too. At NVIDIA’s GTC 2024, inmind.ai reports that its lead robotics engineer, Elie Youssef, co-led a hands-on lab on custom robots using Isaac Sim and ROS2. Teaching people to simulate a robot is a revealing public demonstration of the expertise the company sells.

The same puzzle, made of paper

Now exchange the factory for a research institution. UN ESCWA has extensive multilingual document collections, including material connected to the Sustainable Development Goals. The difficulty is finding and organizing relevant information across that collection. inmind.ai’s Istinara case describes machine-learning classification and relevance ranking, with search and filtering across languages.

UNICEF’s problem sits a little further along the document’s journey. Reports need consistent quality and appropriate treatment of sensitive material. inmind.ai built an automated checking tool that applies predefined criteria, flags problematic content and presents quality scores. The distinction matters: finding information and evaluating a report are separate jobs, even when both involve text.

These cases make the company’s range less mysterious. Whether the object is a vehicle component, a mobile robot or a report, the assignment begins with particular rules. The software has to understand enough of those rules to help the person responsible for the result.

Engineering has a physical footprint

Founded in Lebanon in 2018, inmind.ai identifies Bechara Al Bouna, PhD, as founder and CEO. Its current LinkedIn profile lists Greenville, South Carolina, as headquarters. Announcements document expansion to the United States and Qatar in 2023, followed by a UAE and Cyprus expansion described in its company timeline for 2024.

One earlier detail is more illuminating than the map. Its February 2022 Mkalles office announcement describes 1,300 square meters equipped to develop, simulate and test robots on site. That is a substantial physical commitment to work often described in abstract language. Somewhere, the simulation must meet an actual machine.

Its university-linked academy adds another dimension: developing people alongside systems. Company materials describe training and internships, while its published values emphasize collaboration and adaptability. These are useful clues to how it presents its working culture, rather than a substitute for hearing directly from employees.

The work between the platforms

Commercially, inmind.ai belongs in custom engineering and consultancy. Organizations bring a process; the company designs and integrates software around it. Microsoft, AWS and NVIDIA appear as technology partners. Its position is between the technology platform and the organization that needs a functioning application inside existing operations. In this kind of engineering, the specification is part of the product: it determines what success can sensibly mean.

“inmind.ai was able to support us, adapt quickly and guide us through the obstacles in a very flexible manner.”Daniel Neumeier, BMW / company-published testimonial

The meaningful alternatives are an internal engineering team, another specialist consultancy, or a standard application that already fits the task. Custom work makes sense when the constraints justify it. Accurate models, representative data, clear review criteria and workable integration remain essential; a simulation detached from physical reality cannot settle a packing decision.

That is the useful question to take from inmind.ai’s portfolio. Where does work get stuck, and can the obstruction be described precisely? Sometimes the answer involves autonomous machines. Sometimes it involves thousands of reports. Sometimes it begins with a tray that needs packing properly.