Updates from the Tinyverse - July 2021

The HOT-G team has been working hard to evolve Rune and the Tinyverse since our Tiny ML Talk at the end of May.

There has been a lot of progress made since then, but here are some of the highlights.

Rune Core

Our core Rune tooling consists of a compiler for turning an ML pipeline into executable code and a runtime that mediates access between the outside world and a Rune running inside a WebAssembly virtual machine.

More complex Pipelines

We now have full support for more complex pipelines that contain multiple inputs and outputs. You can even provide multiple instances of the same capability type (e.g. two image inputs)!

This unlocks a lot of interesting ML applications like Natural Language Processing and Style Transfer.

The rune will be available on the Runic mobile app soon! Get the app below :)

Cross-Platform Library

We have also started developing cross-platform libraries for embedding the Rust Rune runtime into native applications.

The provided functionality will change depending on what each platform supports, but pre-compiled libraries and header files are available for download from GitHub Releases.

Rune in the Browser

It’s official, you can finally execute a Rune completely inside the browser!

We’ve created a web app and runtime library which supports Runes containing both tf.js and TensorFlow Lite models. Besides static data, audio and video streams from the device can be passed as input to the Rune.

We plan to release this as a package on NPM, but until then…

Runic Mobile

Both the mobile app and the libraries it builds upon now support more advanced pipelines and multiple instances of a capability.

The Team

We are growing our team and building a vibrant distributed and remote-first team. Rust Programming Language is our tool of choice for several core products.

Over the last 4 weeks, we’ve been hard at work looking for new talent to join our team.

So far we’ve done dozens of interviews and screened a large number of candidates, and filled 3 positions:

• An Embedded Systems Engineer for bringing Rune to the world of bare metal

• A Machine Learning Ops Specialist for developing our Machine Learning capabilities

• A Rust Engineer for working on our core tooling

If what we do sounds interesting, check out our careers page or send a message through to careers@hotg.ai.

The Markets

HOT-G is building AI orchestration technologies for several markets where we see edge computing is required. These include (but are not limited to):

• Maritime industry - where there is low to no connectivity, and predictive maintenance has a big need.

• Security of AI IP - where we need to protect AI models using various cryptographic techniques.

• Oil and gas industry - where there is an abundance of data generated in air-gapped systems, and deploying edge models will accelerate time to market.

• Solana blockchain - where we are building the future of computational Metaverse. Rust is a first-class citizen in this ecosystem.

Community

We are officially sponsoring the biggest hackathon in Canada called Hack The North. Our preliminary agenda is looking good, consisting of hosting a tech talk as well as an API workshop. If you are a university student looking for tangible experiences, this is where you should be. Pssst, we are also looking for Rune Ambassadors so if this has you written all over it, drop us a line at community@hotg.ai.

On the same lines, we have been tentatively accepted to do a technical session at ARM Dev Summit where we will commence our 1000 runes initiative. We have plans to host a fun and challenging competition on Kaggle for the best runes so make sure you bring your game face on!

Keep your eyes peeled and follow us as we announce more fun things in the works on our social channels below.

The Future

We are looking forward to the several key product launches and making rapid progress on our enterprise use-cases. The future of TinyVerse is very exciting, and over the next 12 months, we will be building. Some of this includes:

• Support for audio, NLP, and other models

• An additional set of devices out of the box

• Multiple training framework support in Rune

• Hammer Forge - auto-orchestrate your edge deployments

• Deeper collaboration across the community as well as the industry to make Rune the container standard

• and more :)

Join our Community:

Twitter @hotg_ai and @hammer_otg | LinkedIn | Discord | Discourse