Building Teams that successfully implement AI - Hiring for MLOps and Validating the work
More companies are recognizing that they cannot build and leverage in-house artificial intelligence/machine learning models from the very first month of setting up a data science team. With the increase in the influx of venture capital money, global startup hubs are seeing more data science jobs being posted and increasing demand for AI and data science experts.
It is said that when it comes to building a data science team, many companies fail at the first step — creating a job posting. However, it doesn’t imply that the challenge ends there. It can be confusing to hire the first set of data scientists, and hiring managers feel overwhelmed with a lot of opinionated content for “data science hiring,” while the job description of a data scientist may be different from company to company.
While we wrote a blog on how to hire data scientists at a company, we considered asking our friend and one of the world’s best minds in the field about this.
In our recent episodes of Hiring Diaries, we hosted Bob Friday – VP / CTO of Juniper Networks. He is a serial technology entrepreneur. His first tech startup went for an IPO, second technology startup got acquired by Cisco for $450 million in 2005 after which Bob joined the organization as its VP / CTO. And he’s currently leading his third startup in the capacity of a Co-founder and CTO of Mist AI – now a Juniper Networks company.
Mudit Srivastava, Host, Skillspace.ai:
How do you go about finding people around MLOps? While it is very much gaining prominence it is still not at the stage where it should be.
Bob Friday, Mist AI and Juniper Networks:
We’re hiring data scientists here in the US and India. And I would say, hiring in the US and India are slightly different. In the US, it’s more about you’re very competitive in the US trying to hire data scientists. You know, what I found in India, it’s a lot harder. A lot more people in India call themselves data scientists, you know. There’s a little bit of difference between the Ph.D. data scientists, the person who knows how to write something, they know their way around the models and can train them. And then there’s data engineering too. And that’s the other critical part of how do you if you’re building one of these models, a lot of the work, like I said before goes into creating the pipelines. Which is not typically a data scientist’s activity. There are all these different pipelines that have to be built before we even get to the model.
So what you’re doing at Skillspace.ai, I found it perfect to have some way of actually screening data scientists, right some way to filter through all the different noise and get to a good point. Is this person a Ph.D. data scientist who understands all the theory and math behind it? Is this a software engineer who’s taught himself some basic ML algorithms? Or is this a data engineer who understands all the pipelines and how to get all the infrastructure in plumbing built for models? So that’s what you’re doing at Skillspace, it really helps kind of sort that out and I found that’s critical you know, we are over here in the US trying to sort this out in India.
Mudit Srivastava, Host, Skillspace.ai:
Thank you for the kind appreciation of our work Bob. We highly appreciate it.
I want to pick this question that one of our attendees have asked – how do you validate your AI-based startup idea or an internal AI project?
Bob Friday, Mist AI and Juniper Networks:
Well, yeah, and there are probably two parts in the question, one is a startup, and that’s more about what I call product fit market fit. So usually when someone comes to me they’re trying to start something. You see, they ask whether I will fund it or not, but what is the market fit? What are you doing in the market? Why do you think there’s an opportunity?
Mist was around architectural change, right, mature market and we believed there’s an opportunity, there’s an architectural opportunity to disrupt. The other one around product fit is like what are you doing that is differentiated. And usually, when I talk to people who are trying to do some startup, and they’re claiming they’re doing some AI, I ask what human behavior are you trying to emulate if you’re claiming to do AI?
What human task are you trying to build do on par with? And then the next thing I usually get down into is, like you know, describing to me what algorithm you’re using, and usually, it’s not one algorithm, like, say, when you’re building something on par with a human. It’s typically not one algorithm. There are usually a couple of different algorithms.
If you’re making an autonomous car there’s not one magical algorithm that’s driving that car. There are many other little algorithms in that car to get that car. Your autonomous going me dry conditions are a little bit different. You know, you could build a whole startup on image recognition. If you have gotten good at convolutional and recognize, but even there if you’re looking at MRIs and trying to do that, it’s not usually one algorithm is a couple of algorithms. So that’s, that’s usually when people come to me with AI startup concepts is break it down into market fit and then product fit- what AI? What human task are you trying to automate? And then I started digging into what algorithms you know, what math, are you using leveraging here?
Liked the conversation?
This is an excerpt from a recent episode of Hiring Diaries with Bob Friday, VP / CTO of Juniper Networks and Co-founder of Mist AI – a Garner MQ Leader. Our entire conversation covered several topics, including:
In Bob’s journey of doing an IPO and two big company acquisitions, how has raising money changed over the last 20 years?
Theory and the product behind Mist AI.
The ground reality of AI implementation. Is AI overhyped?
Hiring the first set of data scientists at an organization.
What should be the difference in approach while hiring data scientists at startups versus an enterprise?
Insight into AIOps and its recommended hiring practices.
Parameters to look for while hiring data scientists and MLOps professionals.
Key levers that took Mist from a blank sheet of paper to the Gartner Magic Quadrant leader in just six years?
Starting up a tech venture in today’s world.
To view the entire discussion, visit here – https://skllspace.ai/hiring-diaries-204/.
More about Skillspace.ai and Hiring Diaries
The goal of Skillspace.ai is to deliver better experiences for end-to-end technical assessments with scale. In our platform, you can hire tech talent through coding assessments, AI challenges, take-home assignments, 1-on-1 interviews, and much more, all in a highly condensed and simplified manner. The platform is trusted by a range of early-stage startups to Fortune 500s and is built by DPhi.
DPhi is a global community of 100K+ AI enthusiasts from 140+ countries. The initiative aims to democratize data science learning and build an ecosystem to engineer the future of AI. To know more about DPhi, you can visit here. To know more about Skillspace.ai, you can check out this page.
We have created Hiring Diaries to support the engineering and talent acquisition communities that have trusted us so far. The goal is to create a series of live discussions with the top 1% of talent and technology leaders who share our passion for giving first. Hiring Diaries covers various topics regarding scaling technology teams and hiring the best talent. You are welcome to attend any of our upcoming events if you wish to learn more.
