Intelligence at the Edge: Introducing baeIoT

Introducing...baeIoT

The first ever post comes straight from the city of Bethesda, found in the blue crabs state of Maryland. Bethesda is roughly seven miles from the seat of the US Federal Government - Washington in the District of Columbia (DC).

The Federal system of government is arguably the most representative governance model. The US federal government version, from my understanding, is a hybrid of a Republic and a Democracy. Looking at the evolution of governance in the US it appears to me that it has evolved from the centralized power model of aristocracy then to meritocracy and eventually to the decentralized/federated model of democracy.

Similarly, computer technology has evolved from the days of thin clients (centralized model) then to thick clients then to client-server then and eventually to peer-to-peer computing or federated computing (decentralized/federated). This blog will therefore constantly draw parallels in my "baeIoT" work with workings at Capitol Hill!

Analytics on Edge Devices

In my Computer Science Master's studies I fell in love with distributed databases, specifically Apache Cassandra Database (and Facebook developers' Greek mythology motivation of Prophetess Cassandra: seeing the future and not being believed).

Suddenly, I developed a fascination and positive intuition for any technology distributed, decentralized, federated or peer-to-peer. The three that caught my love the most were: Distributed Ledger Technology (Blockchain), Decentralized Cloud Computing (Edge Computing) and Decentralized Artificial Intelligence (Tiny Machine Learning.); I had just started the St Peters IoT Makerspace, the first IoT Makerspace in Zimbabwe in 2016.

Lo and Behold the baeIoT (bae of IoT) moniker was birthed which represents these four technologies I would commit my love to: b-blockchain, a-artificial intelligence & ML, e-edge & fog computing and IoT! This blog is therefore the baeIoT's sojourn into the world of Intelligence at the Edge or Federated Intelligence using baeIoT technologies.

Cloud computing has been one of the key enablers of Data Science as it allows data practitioners to focus on solving business problems whilst delegating data storage and management to cloud service providers. However, in a third world country like Zimbabwe, where I come from, Cloud computing is a luxury. This because of constant power outages as well as Internet bandwidth, reliability and costs challenges.

This has forced projects undertaken in the Global South to stretch and deploy edge analytics on-device or cloudlets in offline-first or edge-first architectures. Emerging analytics paradigms such as on-device Tiny Machine Learning (TinyML), consequently became popular in environments such as Zimbabwe.

We have lately been experimenting with tools in the field of Federated Learning such as TensorFlow Federated framework at University of Zimbabwe. In earnest this adventure was nothing short of an academic adventure. However, developments such ML-specific Neural Processing Units (NPU) and innovations such as the AI-capable Cortex-M55 and Ethos-U65 bring makes the prospect of end-point intelligence a reality.

In conclusion, it is the accessibility of technology or as Wikipedia puts it "democratization of technology" that fuels projects we will be showcasing here on Disruptive IoT. Additionally, developer tools, that collaborative and democratizing in nature, will also help us to focus on problem solving, perfect our thought process and to share our work. We hope to interact on these platforms we will utilize: DAGsHub and Edge Impulse