Machine learning
harness the power of ai
I am a director of Octofox, an A.I. company in the U.K. providing machine learning and data science consulting to international clients in the biotechnology, engineering, power and service industries.
I specialise in deep learning and planning machine learning projects and have expertise in productivising machine learning solutions on several major cloud platforms.
My goal through consulting in machine learning is to have a positive impact on the world by harnessing the power of AI.
If you are not familiar with what machine learning is, or you think it’s only for big tech companies, you can read an article I wrote about how machine learning can help you achieve your goals here. You can also read an introduction to artificial intelligence that I wrote here or watch a webinar I did with a colleague at Sonat Consulting on how to plan machine learning projects and avoid common pitfalls here.
Read on to learn about some of the machine learning consulting I do. I also used machine learning in my academic research, which you can read about here.
Machine learning consulting
Deep learning for equipment health management
Monitoring the health of critical machinery and responding quickly and efficiently to avoid equipment failure is one of the focus areas in developing intelligent ships and power stations. It saves time and money and increases safety and security. Machine learning is key to its success.
Monitoring is done via hundreds of sensors that measure physical properties, such as temperature and pressure, of different components. These measurements are taken continuously and produce enormous volumes of data. It is not possible for humans to go through it all. This is where machine learning comes in.
I developed machine learning models for monitoring engines onboard marine vessels and in power stations across the globe for an international world-leading engineering company. The models, based on deep neural networks, detect anomalies in real-time, classify faults and predict the remaining useful life of key components. I also explored the use of deep reinforcement learning for optimising engine performance. This was a large project that I worked on for over 2.5 years.
Machine learning consulting
Databricks and Spark
- Everything is in one place: the data, the SQL queries, the notebooks, the clusters.
- You can work in SQL, R, Scala or Python and combine them all in the same notebook for completing different tasks.
- It is built on Apache Spark, an engine for data analytics that allows data processing on a huge scale, which makes everything extremely fast (I use PySpark)
- You can spin up different clusters tailored for different parts of your analysis pipeline.
I’ve been using Spark and Databricks extensively for the last 1.5 years while consulting in machine learning in Bergen, and I co-led a training day at Sonat on it.
Machine learning consulting
NLP and chatbots
People increasingly want to interact with machines using natural language. But this requires machines that can understand them.
Natural language processing (NLP) is a huge area of machine learning. Chatbots are one exciting and increasingly popular application of it.
I co-led a training day for Sonat on the use of NLP and chatbots.
Machine learning consulting
Planning ML projects
Machine learning can save you time, energy and money and it can accomplish things not possible for humans alone.
But machine learning projects also require careful planning and a well-defined goal to avoid the trap of meandering through a never-ending wilderness of data.
To stay the course and achieve the goals you set, you need a robust framework for iteration and testing and expert knowledge in how to evaluate the output and value from your models.
I enjoy planning and have several years of experience planning machine learning projects and developing processes to keep them on track. I also planned research projects, as well as large experiments, while I was in academia.
Watch on YouTube: how not to fail at machine learning.
Machine learning consulting
Generative adversarial neural networks (GANs)
GANs are at the cutting edge of AI. They allow a machine to create new data:
- Create realistic images
- Create a picture from text
- Translate images from one style to another (e.g. day to night, satellite to map)
- Improve data quality
