You have found a blog on the real-life aspects of machine learning intensive products
There are huge amounts of materials on machine learning, data science and artificial intelligence on the internet. What I find lacking is information on the practical aspects of ML-heavy systems such as how to develop them efficiently, how to automate the time-consuming routine tasks in the modelling process, how to make sure the model is behaving correctly after it is live and so on.
I hope to make the blog useful to other software engineers who work in the machine learning area as well as data scientists who would like to know more about the ML product development outside of modelling. The content will be a mix of the emerging best practices and my own thoughts on how to engineer ML products and shape ML infrastructure for the rapid development of new products.
I currently work as a software engineer at Klarna where I develop distributed systems for real-time credit risk assessment and fraud detection. In our team we mostly use Java, Scala, plenty of Kafka and some Hadoop-y stuff, we continuously deliver our containerized apps to AWS. Well, technically it's not us but Jenkins who does the hard work but we like to be involved. Credit risk and fraud are some of those areas where even the best hand-crafted rules are no longer competitive and machine learning capabilities are vital. This is where I started working with teams of data scientists and became interested in the engineering aspects of their work.
Before that, I worked as a consultant for Deutsche Bank, concretely on their reference data services in the area of Corporate Action Event processing. That was a Java EE shop with all sorts of enterprisy things like Weblogic clusters, SOAP over JMS, ESB, BPM, tons of documentation and some TOGAF practices.
And before that, I was assigned to a project for T-Mobile Germany where we made a portal for managing mobile phone customer subscription settings. That was my first job starting in 2008 where I learned a lot about using Java for real projects. That was where I saw the most SOAP services, echoes of NGOSS!
Between T-Mobile and Deutsche Bank projects I helped out Exigen Services, the consultancy firm I was employed at, with some pre-sale activities, mostly in the insurance domain. I even got to work with SAP stack for a short while and wrote a few lines of ABAP code.
Apart from work, I spend a lot of time with my better half Elvira, exploring parts of Stockholm and its surroundings or visiting neighbouring EU countries.
There's usually an online course or two I am taking to keep up-to-date with ML and SWE stuff, consider me a believer in the lifelong education. Thanks to Coursera, edX, Stanford lectures on NLP and CV, and now fast.ai, keep 'em coming! I enjoy learning new programming languages too and looking for an opportunity to learn myself some more Haskell for a great good. Go has its nice sides, I also like Erlang, Scala and Python and would use them depending on the task.
Disclaimer: I am not necessarily an expert in many things I will blog about, rather a learner hoping to get better by writing.
Have questions or comments about the blog? Or maybe you are a non-profit looking for a software/data engineer volunteer?