#12 – Phil Coachman – Principal Data scientist demystifies “Data Science”, the future of Data Science and a really cool ML project
If you are in the data science field or aspiring to be in the data science field you are going to love this episode. In this episode we talk with Phil Coachman who works as a Data Scientist/Principal Cloud Solution Architect in Microsoft. We talk about Phil’s journey starting with the Bing, Xbox and Windows 10 team to his current role. We also demystify “who a data scientist” really is and what it take to be one. We talk about a really interesting and cool Train derailment prevention project using AI, Machine Learning, IoT sensors and IOT edge which is currently in production and has been deployed to 4 continents. Phil also gives you 4 ways on how you can keep up and learn everything you need to know about Machine Learning.
Enjoy listening !
Phil’s journey on how he got to where he is today (01:52)
- Phil started off working with a project codenamed “Live search” (now called Bing) 12 years ago- and he had the opportunity to work on tools like Hadoop, Spark and other big data tools and capabilities.
- Working with big data processing with 100’s of PB for Bing, Xbox and Windows 10 – he understood the relationship between data and insights and naturally progressed into a Cloud Solution Architect specializing in the data science field.
Demystify the role “data scientist” (03:59)
- Most people who have the title of Data Scientist are actually doing Data Analyst roles where they work on Excel. The key is a “scientific method”, creating a null hypothesis and going about to prove it you are not really doing data science. You could be doing data analytics or data engineering or data mining.
- The future of Data Science as Phil’s perceives it – Data analyst are gradually moving towards data scientists as tools become easier for machine learning and deep learning.
- Excel, Power BI makes it easier to do data science without even realizing you are doing it. Tools like DataRobot, Julia, and Azure automated Machine Learning makes it even easier to figure out a what is a good model without having to know how the actual algorithms work behind it.
Interesting projects that Phil has worked on (07:57)
- Train Derailment Prevention – Customer who deals a lot with Trains had a lot of trains across the globe with 1000s of sensors across the train like heat, vibration etc. Their goal was to monitor the trains and also be alerted when a train derailed (for trains running in remote location like rainforests it will be harder to find derailed trains).
- POC was using Raspberrry Pi and a toy train set inside their lab. A model built in R and get into Azure stream analytics and get alerted when a train derails. The POC was a success. Big Edge devices were placed on the train. The production was then deployed across the globe and they used this to sell the solution to other train manufacturers and vendors across the globe.
- As a phase 2 – they are picking up sound frequencies from the wheels to predict if a bearing is going to fail – so that they can proactively fix it and thereby reducing train delays.
- The resource constraint also meant that the solution had to work with 2G network connectivity which was overcome using IoT edge to bring AI closer to the source as possible.
Phil’s advice to anyone who is starting off in the Cloud Solution and Data Science space (14:26)
- Holistic understanding of the other pieces involved in a solution is critical to deliver a quality solution – example Networking, Governance, Security, Identity understanding is critical even for a data science project.
- Collaboration is crucial for a success solution architect to leverage the strengths of the team.
Phil’s advice to keep himself abreast of all the changes happening in the cloud world (16:42)
- Top 4 ways he keeps himself up to date:
- TowardsDataScience – site he follows this – articles from students, practitioners and professors
- MachineLearningMastery – similar to the above site
- ACM – Latest from the university and research papers.
- Kaggle – ton of data sets and new challenges. It helps you take out the tunnel vision and take a broader picture.
- For work life balance –
- Phil blocks time on his calendar for learning and keeping up with the latest updates.
- When he is travelling he uses hotel hours to keep up with new things.
How to get in touch with Phil (21:07)
- LinkedIn – https://www.linkedin.com/in/philcoachman/
www.cloud-monk.com
. Show Notes Available at https://cloud-monk.com/2020/02/01/ep12/.