A.I. product life-cycle management utilizing deep learning applied to computer vision
Allegro is the brainchild of Moses Guttmann, Nir Bar-Lev and Gil Westrich. Back in 2014 developments in AI and more specifically in deep learning, were making headlines regularly in academic circles. These events provided a glimpse into how this technology would revolutionize computer science in general and more specifically bring about a paradigm shift in computer vision. The team identified three key realities:
AI was a fundamentally different software product paradigm. The abundant ecosystem of tools, infrastructure and platforms available for traditional S/W products wouldn’t suffice any longer. To support the scale-up and roll-out of deep learning-based solutions by product teams, new tools would need to be created.
AI was a learning-based paradigm, but all the work on AI was essentially creating “learned” [static] solutions as opposed to learning [dynamic] solutions. The learning always ceased when products were deployed, although AI has the potential – in fact one of its key strengths – to continue learning post-deployment.
Real-time computer vision solutions require processing / decision -making to happen at the edge. AI, and more specifically deep learning, is a very computationally heavy technology. The problem then becomes: How to produce results with limited resources in real-time? The only architectural solution that is device and H/W platform agnostic is one that delivers a uniquely tailored solution to each edge device and its uniquely specific environment. As such, the S/W solution to the problem carries the highest ROI and widest use-case applicability.
With these realizations, the three teamed up in 2016 to embark on a journey to address these challenges and opportunities with a vision of enabling every organization to leverage the full power of deep learning to build solutions for a better world.
From the get-go, the founders understood that any best of breed platform requires a combination of deep domain-expertise as well as superb software engineering in relevant areas. So the partners set up to build a 360-degree team in terms of the talent surrounding the problem statement. They brought in key talent in deep learning, high performance computing, backend, DevOps, embedded development and product management.