How to Hire a Great Lead Data Scientist

Whether you're trying to hire a Lead Data Scientist, or wanting to be one, I think this is a good framework to keep in mind when ascertaining the skill and tenor of a Data Scientist. With that said, let's dive into four (4) quick hitters to try to solve for when you're trying to hire this Data Scientist.

1. Scope and Define Problems

They should exhibit a pointed curiosity and great skill to efficiently pepper questions to help you define problems that will impact the business in metrics that matter to the business (e.g. Revenue, User Acquisition, Feature/Product Dev).

They should restate and summarize the problems. If you don't know, they should show an eager desire to help to try you define some possible low hanging fruit.

As they go through the interview they should be continually revisiting with more questions/ideas/suggestions around the problems you initially scoped together.

2. Watson, Know Thy Data

They should probe the current state of the data you have and that will support the scoped problems. The Cotton Eyed Joe (Where Did It Come From, Where Did It Go!) of the data to its structure, formats, feeds, frequency, amount, API Logs, Clickstream, Salesforce, Marketing data, etc.

Their job as a Data Scientist is 100% dependent on getting data. They should know how to get data (self serve) and if not...

3. Compute, Platform, and Support in Place: DevOps/DEs/Tooling

They should seek to understand the systems in place to support them. e.g. Dashboards, Jupyter Notebook server, Cloud Compute (how to get it), ML Models Server (Docker, ECS/Kubernetes, CI/CD pipelines, Flask/FastAPI), Ad hoc data analysis, batch Reporting.

They should know how data is accessed through all of these, if not, that's a red flag that they will need to be fed and held hands hoping you'll know what to do.

Here you should start to see them connect what they need to learn and where their current skill set in tooling applies.

4. Meeting Those They Will Work with The Most

Last big point is this candidate should either have figured out (and confirmed) where the role sits in the organization and at this point I expect them to ask to meet the people (or a representative) of the org(s) they would most closely work with. Almost always this will include someone from Product. You're VP of Product, the Director of Product, or a strong Technical Product Manager who has a strong understanding of the business to be able to have a great solutioning session with the Data Scientist candidate. There probably could be another person from another department, likely defined from the problems scoped in point #1, e.g. Marketing or Sales if building a Marketing Engine or Pricing Engine. Or perhaps someone from Data Engineering if there seemed to be some issues with Data Quality. If building and deploying ML models to production they should likely ask (and speak) to someone who would help support that venture, e.g. DevOps, MLOps, ML Engineer.

If you haven't already stated and scheduled these other people to interview with them I do expect this Lead Data Scientist to ask to meet with them. You could leave it open ended to see if they bring this up, but on the other hand, it could reflect poorly on you that you didn't already understand where the role would sit and who the person in question would work with.

Early Stopping

If they weren't able to lead a session in scoping and defining problems that will make an impact to the business -- you can almost sure they won't impact the business in a positive way. Really hard to bring value to a business if they don't have any interest or idea of how that happens. The Data Scientist's #1 job is to understand the business and bring a data, product, and design driven approach to adding value to your company. If they weren't able to do this perhaps you may have a Jr Data Scientist or Data Analyst role for them. At this point if you had something you liked about them, then a simple tech screen for something like SQL or Python may suffice for the lower role.

The Other 20%

You should figure out what they want to do, what interests them. A lot of times Data Scientist will be voracious learners and thus may not want to do the same thing over and solve the same problems over and over. So need to figure out if the problem space is one they're really interested in, especially if not already apparent from the interview thus far.

You need to know how willing they would be to pitch in to help any problems they noted upstream, e.g. ETL, ensure quality API Logs (e.g serialized JSON), be able to work with Front-End Engineers to get good clickstream data, build a Docker container, build a Dashboard, build a Data Model to support a dashboard, etc.

You need to know how self directed they are. How they learn. Where do they go for learning. Do they know who the leaders are in their industry and understand the tooling landscape? They will be picking tools as they work and should constantly be on the lookout for ways to improve the efficiency and effectiveness of their workflows.

Do they explain technical terms in ways you can understand?

Do they seem pleasant to work with? Do they show humility? Are they short, obtuse, or show any elements of pride? They should be more concerned about the data and what is right than being right or seen as smart.

Final, Parting Shot

Ultimately, this hire will help shape the business in a dramatic way. Be sure you are intentional in finding Data Scientists who are willing to get their hands dirty and adept in scoping the projects that add value to the business. Their hands will be instrumental in helping shape and guide the rest of the organization.

#datascientist #dataanalyst #hiring #mlengineer #devops #dataops