Production grade perception systems require high quality foundations
In the perception space, there is an abundance of software tooling that is extremely well suited to a research environment. However, many of these solutions fall short in production. Common reasons include:
Poor software quality: inadequate testing, bad design principles, low quality/missing documentation, inadequate error handling and observability, difficulty in extending and adapting.
Ecosystem lock-in: for example, not everyone can run a Python-based stack in production
Bloat: large dependency trees make integration difficult and a perpetual maintenance burden.
The “boil the ocean” mentality: one tool to rule them all, trying to do everything adequately rather than one thing really well.
At AFWare LLC, we have many years of experience building production grade perception systems and have suffered through working in codebases that rely heavily on such tooling. Pragmatically, we understand that businesses cannot necessarily afford to devote engineering time to “reinventing the wheel” and “doing it the right way”, even when these shortcomings are acknowledged.
Why AFWare?
Our goal is simple: to make that suffering unnecessary. With over a decade of academic and industry experience, we aim to provide foundational perception libraries and tooling that lets companies build things “the right way”, from the start.
Our solutions are designed to be:
Performant
Maintainable
Adaptable
Simple: adding complexity only when absolutely necessary
Philosophy
Our design philosophy is that perception software should:
Be crafted & maintained to the highest standard by domain experts
Easily integrate into existing projects: keeping the dependency chain light, using standard tooling and “playing nice” with existing build infrastructure
Be predictable: behavior and interfaces should be unified
Well documented and tested
We have many exciting projects in the works. Please watch this space for more information.
To get in touch with us, please use our Contact Us form.