Let's transform your signal acquisition hardware into a cloud-managed edge lambda service
Providing extensible event handling for latency-sensitive hardware signals can be challenging. Our Rust-based software agent's embedded Lambda service runs low-latency Lua and WASM lambdas, enabling intelligent inference at the source of the signal.
Let's bring your compute to the signal
Bringing compute to the source of the signal can be challenging.
You've tried lift-and-shift, like Python based inference using 10 GB Docker images on resource-constrained latency-sensitive edge hardware, and you've hit the wall.
It's time to operationalize for the edge.
Our Rust-based software agent embeds a Lambda service that runs low-latency logic written in Lua, JavaScript, or WASM , enabling inference to inform control systems right at the source of the signal.
Proudly Serviced
Here’s How We Can Help
Kick the tires by quickly developing lambdas in Lua or JavaScript that directly respond to hardware signals, right from our cloud console.
Then develop beefier lambdas in Rust or Go that compile to WASM, for more complex workloads such as inference powered by ML or AI.
Cloud Console
Develop lambdas in our collaborative web console
Lambdas
Run logic that responds to events, right at the source of the signal
Lambda Triggers
Write logic that generates the events that Lambdas react to
CLI
100% of configuration can be done via the CLI or CI+CD GitOps
Rust Ecosystem
Leverage inference crates for Onnx, Tensorflow, and more
High Performance
Our agent is Rust based to help your intelligence keep up with your signals
How We Help You Get Started
STEP ONE
Inquiry Form
Click Get Started to fill out an inquiry form describing your goals.
STEP TWO
Schedule a Call
See a demo, and explore options ranging from self-service to paid pilot.
STEP THREE
Install the Agent
Start building your first Lambda that deploys to your on-premise agent.
Edge Intelligence Enabler
Hi I’m David, CEO of On Prem
As a seasoned software professional with decades of start-up and Fortune 500 experience, I know analytics.
In the cloud, analytics is performed on big servers using distributed clusters of Apache Spark or DataFusion. I worked on several of those, including the Spark Service at IBM Analytics, and the GPU-accelerated Theseus SQL processor at Voltron Data.
Intelligence in the cloud is fed STALE DATA that has to be brought in from the edge, sometimes at great expense.
Fast forward to today. Inference is moving closer to the source of the signal, where insights can inform control systems without requiring a slow, expensive, or unreliable connection to the cloud.
On Prem focuses on Lua, JavaScript, and WASM, three language runtimes that are proven to deploy reliably over the air or be embedded, and are uniquely able to orchestrate performance-sensitive code.
Frequently Asked Questions (FAQs)
Some of our most frequently-asked-questions:
Q. What are some of your licensing options?
Getting started is free. Just download the agent, log into the web console, provision an API Key, and get started using the examples at docs.on-prem.net.
If you're planning to operationalize a Python inference workload and don't know where to begin, you could consider a one-week professional services engagement to explore porting it to Rust+WASM.
If you're an equipment manufacturer interested in a white-label license of the On Prem cloud console and its API endpoint, to bring cloud-managed over-the-air extensibility to your customer's device fleet, ask about our paid pilot programs.
If you perform device management at scale, you might consider licensing the On Prem control plane and integrating it into your own. When used with the On Prem agent, it has capabilities with respect to Prometheus middleware and Redfish middleware, and can leverage the Lambda and Lambda Trigger (k8s robotic control loop) design patterns to realize BMC device drivers, Redfish action hooks, and attaching reactive performance-sensitive logic such as inference or downsampling to high-frequency signals right at their source.
Q. I'm new to WASM. What's its basic value proposition for inference?
The Rust ecosystem at crates.io contains a rich assortment of high-performance and memory-safe libraries, ranging from Onnx and Tensorflow runtimes, to image, video, and audio processing libraries. Rust compiles to WASM, and then On Prem is able to deploy it over the air to your edge software agent.
This keeps your workloads lean and fast, so you can avoid those insanely unmanageable 10 GB Python Docker images.
Q. How can Release Versioning be achieved with your SaaS?
Lambdas can be developed interactively via the cloud console, and targeted to run on a software agent running on your laptop, or specialty equipment in your test lab designated as a development environment.
Assets can then be exported as script files plus JSON or YAML metadata, and checked into Git.
Continuous Integration tests can rebuild your environment from scratch for each merge request, using our CLI to import your assets.
Q. Why is your agent written in Rust, not Go?
I know of a Go-based software agent that tried to keep up with a 4-channel 24-bit A/D converter, and topped out at around 22,000 samples/sec (almost 20x too slow). That's when I learned that low-latency signal processing is at odds with the garbage collector languages.
Q. Why is your agent written in Rust, not C++?
Rust provides the agility required to execute projects swiftly, while delivering performance in the ballpark of C++, while delivering best in class memory safety.
Let's get started adding cloud-managed self-service extensibility to your speciality hardware
Click the button below to get started