Ask Me Anything: Stuart Davie

@ioannis.doukas has just beaten me to a similar question!

With an increasing number of businesses applying AI/ML to their commercial decision making in similar sorts of ways, can you imagine a sort of “convergence” in how businesses of a given type operate as a result?

What consequences do you think this might have on data science as a discipline (if any)?

Hello! A lot of data scientists (like you) make the transition from academia to commercial data science. What was it that made you want to leave your role as a university researcher, and how did you find the transition? What do you enjoy the most about your job at the moment? Similarly, do you miss anything about academia?

As a Data Scientist we might be actively working on a project like Inventory optimisation, customer segmentation etc. My question is “How to keep up with ever-growing field of Data Science, Machine Learning, Deep Learning, Computer Vision etc while working on our projects?” 

Hey Stuart, what would you say is the best thing about being a Data Scientist at Peak? And why is it different to being a Data Scientist anywhere else?

What’s your main in Mario Kart? 

Hey Stu! 
R or Python?

Hi hi Stu!

With your role more focused on strategic work than day-to-day data science, how do you keep your skills and knowledge honed and up-to-date?

Follow up Q - is this different to how you kept up-to-date before your step from day-to-day to strategy focus? 

@ioannis.doukas 
 

When I moved into a machine learning focussed role in 2013, the name Data Science hadn’t really taken off yet. We were working with ANNs and GPR that we coded ourselves in Fortran. Some groups were using version control but not all. When I moved across to industry and started working on NLP, BERT models didn’t exist, and Hugging Face only just started. So one thing about this industry is how incredibly fast it moves. The field is vast as well - Data Science can cover everything from fairly deep research roles to very commercial focussed roles.

To stay relevant into the future, I think it is important to stay curious, and keep learning. Having a broad base of knowledge and being adaptable means you can focus on a particular area of data science for a while in your career, but still pick up what you need to move to something more relevant in the future. Another important skill is being able to determine what the most important problem to solve is, so you can make an impact. This isn’t always obvious, and can take time to develop

@jloxh 

Yes, I think so, but I don’t think this is unusual. Businesses innovate, and when it is successful, the rest of the industry follows. In the ‘00s, the Netflix recommender algorithm was revolutionary and the Netflix prize was a big deal - now most websites have a recommendation system of some sort and there are point solutions for all sorts of other things. But this doesn’t mean progress in this field has stopped by any means.

Every business is different, and I don’t see using data (information) to do new things, make better decisions, and gain a competitive advantage going away any time soon.

@sarah 
 

There were a few push factors for me. I felt the research happening in academia was slow. I felt that there was a lot of siloing of skills within research groups and uni’s, as opposed to broad collaboration across them. There are problems with the funding model, and the thought of eventually being a full-time grant-application-writer didn’t excite me.
Pull factors were: the exciting research coming out of private companies. Being able to have a bigger impact, faster. Joining a fast growing startup meant there was exciting potential growth options that would test me.

I found the transition ok! There was a lot to learn, but to me this was part of what made the move appealing. I was the most experienced data science when I joined Peak, so had to quickly learn about everything from NLP to predictive customer analytics in order to support the business. I spent my time at Manchester University working in Fortran and bash scripts, so had to pick up R and Python pretty quick as well. 

I do miss real research. This is something we have been able to have bubble along in the background at Peak. We have hosted a KTP partnership with Manchester University, are sponsoring an upcoming PhD, and have hosted about a dozen MSc students. But Im looking forward to growing this function in the team soon

@saurabhsingh 

1. Success is different for different people. Some people earn a lot of money for low impact roles in big companies. Some people have a huge impact in a field that they care deeply about for low money. Some people are named on lists of most influential people in the industry at a young age; others work from home on niche, interesting problems then log off early every day. Good depends on the job at hand, but there are some values like being Curious, Driven and Responsible that are always helpful

2. I don’t think there is a good answer here beyond trying different things out and self reflection! There are a lot of transferable skills across the roles, so getting exposure to one area will help with others. I worked with a talented data science manager in the past who stepped out of the team management role back into an individual contributor one. The exposure she had to leading a team meant she understood some of the things that might have been going on behind the scenes, which meant she knew how to be more impactful in an individual contributor role. There are frameworks for some of this analysis though, such as SWOT.  The only other thing I would add is that there isn’t a need to rush. Data Science is an interesting career, and focussing on solving enjoyable problems and developing your own skills and experiences (breadth and depth) will open doors.

3. There are huge differences between a project manager and a team lead role, although at Peak some of the responsibilities overlap (part of this is the nature of being a growing business). A great project manager knows how to lead projects success. This includes scoping, setting project expectations, external stakeholder management, calling in reinforcements when required, and other things like this. It can involve difficult ‘external’ conversations. It involves leading people, but for the success of a project. Being a great people manager means caring about the long term success of the people in your team. It involves helping people develop into the best they can be, and supporting them shape their career. It can involve a lot of difficult internal conversations. Many people who interview for team lead roles at Peak conflate the two, but they are distinct.

4. Imposter syndrome is very common in our field, and there are a number of ways to address it. I wouldn’t do it justice here, but am happy to follow up if you want. Personally, I have been fortunate to have not been in that position (yet). I have spent a lot of time in groups where I am clearly in a developing role, or in a position where I have unique skills or experience. I think being open about your strengths and weaknesses, and adopting a growth mindset, can be helpful, but it is a big, tricky issue.

@amysharif 

Tough question, given how many businesses it spans! I think there are a few common challenges that overlap, and are more prevalent in certain businesses. These include:

• Smaller businesses: Insufficient data or data maturity to work on interesting things. A data scientist will often have to do a lot more data engineering or BI-style reporting than they expected
• Older businesses: Siloed systems. Politics. The lack of a communication layer between the data science team and internal customers/end users meaning there is little buy in, and the wrong things are built. Challenges getting things ‘into production’
• Banks and similar regulated industries: Regulatory overhead resulting in very slow moving projects, and importance of interpretability limiting the breadth of techniques that can be used to solve problems
• Service-heavy companies / consultancies: Conflict with sales teams or account managers who don’t quite understand what is possible/not possible using data, and who want to promise definite timelines and outcomes for things that are entirely dependent on data and processes that haven’t been explored yet. It is Data “Science” because we can’t actually be sure what the outcome is going to be, and we need to experiment to get there.
• Any company if you are part of an under-represented group: Representation and inclusion. Data Science as an industry doesn’t have great diversity, and this can result in strongly biassed ‘default data scientist’ mindsets, and a general lack of inclusion. I recently read that women leave tech at a rate 45% higher than men. Some might read this as a challenge for tech to retain women, but I read it as the consequence of the challenges women in tech face to feel valued, respected, and included.