We are a team of seasoned software and machine learning engineers, the team behind MarshallAI. We are committed to build robust ML/CV solutions to guarantee safety at work and security on site. We’ve been on the journey for almost 5 years now.
The journey started with interest in applying deep learning to solve real world problems and willingness to bring more safety to the world. Since then, we’ve been focusing on developing the world’s leading production grade deep learning systems. The platform has been used for e.g. CCTV mass monitoring, command center solutions, workplace safety and PPE monitoring and police chest cam solutions for our partners and customers in the defence, security and safety industries.
In retrospect, the last year seems to have been a real game changer in the industry. Deep learning based defence, security and safety solutions started to take off globally - and we are not talking about the general AI hype. Def/Sec/Safety industries have always been in the forefront of implementing new technologies. And the hype was here about 5 years ago when we started. It’s the second wave we’ve seen starting to rise last year.
There are two obvious reasons for that. One, the market has matured and the driving use cases have been found and validated. On the other hand, so has the deep learning technology and especially the experience of engineers, so currently we really know how to use this powerful technology in production.
To illustrate this change of momentum, we’ll share a typical discussion with our clients from all over the world. We have customers in Japan, Singapore, Germany, Sweden and the US, most of who initially contacted us based on word-of-mouth recommendations made by our existing client or solution partners.
The message and the following discussion goes mostly along these lines:
I heard from person A about your company. It seems you are pretty much doing what we’ve been looking for. Could you kindly summarise what are you exactly doing? What are your core advantages?
Production grade deep learning based computer vision solutions for defence, security and safety industries.
Our key advantages are:
What do you mean by “hardware agnostic state-of-the-art performance”?
We have spent years to make these things work out smoothly. Our MarshallAI based solutions are currently optimised to meet almost any hardware or use case requirement, including specific accuracy, distance, input data quality, data type, inference time and cost from GPU clusters to edge FPGA. The volume can range from multiple real-time streams to terabytes of footage, and the data type from digital to thermal or audio signals while maintaining required performance.
What does production-ready mean to you?
The entire system has been designed to meet requirements of high availability and robust performance such as containerised modular design, testing, monitoring and integration capabilities that are needed in a real production environment.
Are you saying that MarshallAI is already in the production usage somewhere?
Yes. Our solutions based on the MarshallAI platform have been successfully running in production for 3 years now.
Could you give examples, which have been your greatest challenges or findings along the journey?
Surely. The first thing is of course confidence and privacy practices. Our spouses have no idea what we’ve been doing the last years. Seriously, one real example could be the fact that in some of the cases the images, videos, sounds, logs and time-series are not allowed to be moved away from the physical devices of the end user. It causes challenges, as the deep learning models and related algorithms need that data to learn new and improve existing skills. We’ve been working on that for a while now. As a result, MarshallAI now has “federated learning” capabilities, which makes it possible to share knowledge between nodes in different physical locations, without compromising data privacy or security. On the other hand we’ve build tools that makes it possible for local operators to teach the system further and fine tune the performance of MarshallAI.
Well we have several use cases where we’d like to evaluate MarshallAI. Could we arrange a new meeting with all the stakeholders involved, so we can decide on concrete next steps with our business owners?
I really think we are onto something. The market seems to be ready. During the last 5 years we have built a solid base on which to build sustainable growth in future. We are ready! We are here to help you guarantee safety at work and security on site!
We are a Finland based, privately held and independent company with customers on three different continents.
A week ago, an experienced tech business development professional, Marcus, joined our team as the CEO - he’ll introduce himself and his plans for the rest of the year in the next post.
To meet the increasing international demand for Marshall we need software engineers to join on both of our internal teams: SRE & Platform and Deep Learning & Solutions teams. We will provide more details shortly, so stay tuned if you are interested in building world class ML/AI solutions to guarantee safety and security.
In the following months we are also looking forward to announce a few strategic platform partnerships. Hmm.. did we miss something? Oh yes. Our participation in several events is confirmed in both Europe and the US. Our marketing team will soonish publish a separate post about these events!
We’ll keep you updated about details on how to find us and post other events & showcases as the are confirmed.