Arjan Scherpenisse is an experienced Elixir and Erlang developer with
a background in AI. After freelancing as a backend developer for
several years, he founded Botsquad, an Elixir-powered chatbot platform, which got acquired last year and is now part of Enreach, an european telco. Besides coding, he enjoys teaching and transferring his knowledge to the next generation of software developers.
Modern NLP (Natural Language Processing) tasks often build upon large, pre-trained language models like BERT. Neural networks that use these tend to take up a lot of memory, which makes it difficult and costly to scale.
In this talk I present the QnA ninja, a classifier service that recognizes text for for example answering FAQs. Elixir is used to coordinate the classification and training of multiple intent classifiers concurrently. It is capable of scaling by using BERT as a feature extractor combined with distributed Elixir to coordinate pools of Python worker processes.