Ameru Smart Bin: From idea to physical product in record time

Real case studies are rare. All the more valuable are examples where we can understand how an idea becomes a physical product and a viable business model. Ameru AI is just such a case and produces smart waste systems as a B2B solution. Founded in 2021, the company already had paying early adopters in mid-2022. The transition to full-time development and series production was completed by the end of 2024. That is product velocity in action.

The product

Business Case

Commercial and institutional waste disposal suffers from a structural gap between aspiration and reality. Users want to dispose of waste correctly, operators need clean streams of recyclable materials and regulators demand measurable compliance. In practice, however, manual sorting and labeling lead to a high level of incorrect throwing, contamination and loss of value.

Ameru closes this gap at the point of disposal. Instead of hoping for perfect user behavior, the system shifts the sorting accuracy to an AI-supported product. Classification takes place on site and accuracy is continuously improved through real usage data. The result is fewer lost recyclables, less contamination of regulated streams and a clear economic incentive for operators thanks to falling disposal costs and short amortization times. User convenience, operator economy and regulatory objectives are brought together in a scalable product model.

Ameru's target group includes airports, canteens, offices and similar establishments. What I find remarkable is how clearly and distinctly Ameru Business case for these target groups and presented.

Procedure

The way there was deliberately unconventional and low-risk. On the Ameru website a timeline which vividly documents the development with pictures and videos.

• Start as a side project, more of a hobby than a business
• Early experimentation with the core technologies of machine learning and mechanics
• A paid prototype in your own coworking space as a first reality check
• Fast feedback from real use with the option to make quick improvements
• Consistent use of rapid manufacturing methods such as 3D printing and collaboration with manufacturing service providers for metal parts and housings

The decisive factor was not perfection, but learning speed combined with risk minimization.

Product Velocity in use

All four product velocity principles can be observed at Ameru. They do not manifest themselves as a process diagram, but as lived practice along the Velocity Loop from Business, System Stewardship, Engineering and Delivery.

I have already published an elaborated case study on this: The development of the Universal Interfaces follows a standardized template for case studies. Readers can see at a glance what it is all about:

• The Title describes which company is active (Ameru) and which product or which component of the product is involved (Universal Interface)
• The Subtitle describes the added value (easy replacement without tools, future-proof interface)
• Relevant Principles, which are used here (Architect for Flow and Accelerate)
• Relevant Fields of activity the Velocity Loop, which are relevant here (Structure, Design, Commit, Operate)
• The Product image helps to grasp the situation at a glance

Here you can see the picture, the link leads to the full case study in English, next to it an explanatory video.

Further case studies are already available on the website from SpaceX and ZF Friedrichshafen set. Since Ameru will be the main use case for the book, I have already planned the following case studies here, which will be posted online shortly:

• Differentiated stakeholders and measurable value: How explicit KPIs for users, operators and regulators help to resolve conflicting goals and set clear priorities for the Smart Bin.
• Early decoupling of mechanics and ML: How early experimentation with machine learning and mechanics for the Smart Bin mitigates technical risks, stabilizes the later architecture and reduces change costs.
• MLOps in the field: How real use does not become a risk, but a source of continuous improvement by systematically collecting user feedback on performance and feeding it back into development via MLOps.
• Pull-based integration tests for hardware changes: How to avoid integration test queues by triggering integration tests on an event basis and running them in line with the existing cadence.

Conclusion

Ameru shows what is possible when learning is systematically pulled forward and economic reality is part of the development early on. A paying customer replaces any PowerPoint validation. A clear architecture replaces hectic integration. And continuous feedback from the field turns a physical product into a learning system.

Anyone talking about speed in the development of cyber-physical products should focus less on methods and more on such stories. Ameru is one of them.