In this talk, I will examine the landscape of IoT devices and the opportunities in the current Erlang/Elixir ecosystem. As more and more people have an interest in deploying ML at the edge we face a future full of untrustworthy and unreliable tiny data collectors with a limited view of their local environment. I will explore how the existing work in distributed systems combined with resilience and reliability techniques could help minimise this issue. Further, I will examine new frameworks and approaches in Federated Learning within this application domain and conclude with open challenges worthy of further research and commercial development opportunities.