Your browser is currently blocking notification.
Please follow this instruction to subscribe:
Notifications are already enabled.

Challenges Of Building A Machine Learning Team In India

Challenges Of Building A Machine Learning Team In India

Over the past 2 years, most of my meetings with fellow entrepreneurs invariably reach the topic of hiring.

‘How hard is it to hire good tech talent?’ In one word, ‘Very!’

Given that we, at TargetingMantra have gone through the pains and pleasures of creating one the best Machine Learning teams in India, here’s sharing what we have learnt from our experience.

The Basic Tenets

  • Founder’s Job is to Hire

For an early stage startup, your real job, other than fundraising and sales, is to hire key people. Get involved full time and define how you want to shape your leadership team and what key skill sets you need in your team. Focus on what you want your product to look like 6 to12 months from now and build a team for that.

  • Sell your Vision

Your startup is the sum of people you surround it with. Early employees will form the think-tank of your company and will spend a lot of time together. Make sure they believe in your vision and are aligned to dream together.

  • Hire fast, Fire fast. NOT!

DON’T HIRE FAST. When it comes to hiring, take your time. Good candidates wait for good opportunities. Make sure that you don’t take any hasty decisions when it comes to hiring people. If there are some doubts, ask other members of your team to take another round. Whatever you do, DON’T hire just because you desperately need people.

  • Not only ‘Doers’ but ‘Thinkers’

Perhaps the most important quality you need in your employees is the ability to think. Check them for their ability to ask ‘Why are we doing this?’, ‘How will the customer benefit from this?’, ‘Is this the best way to solve customer’s pain point?’

Hiring For Tech

  • Major Concern

The idea of machine learning is fascinating but is difficult to understand. So even though most candidates say that they know machine learning, very few people have already done it. Even if someone has worked on it, the scale of their applications is usually very small. The problem is that the best user models for a data set of less than 1000 will fail for the scale of the big ecommerce companies. Keep an eye open for candidates who keep scale in mind.

  • The First Sifting

The first round is a telephonic conversation. The idea is to figure out if there is a cultural fit and whether the person is genuinely interested in working with TargetingMantra.

Given the activity in the Indian startup scene, good people will get a lot of interview calls. The best way to judge somebody’s interest is to ask them ‘What do you know about us/our work/our company’. You’d be surprised that we end about 80% of our interviews at this stage. 80% of candidates have no clue about what the company does. If someone doesn’t put the effort in going through your website, don’t waste your time interviewing them. Move on!

While looking for a good machine learning candidate, we look for two aspects:

  1. Good with experimentation
  2. Develop production level stuff

You need people who keep themselves updated and are eager to try new things. But experiments fail. The implementation challenges of addressing availability, consistency and scalability don’t arise in an   experiment. It is easy to find someone who works well on one of the two levels. But if the same person does not have the imagination to experiment and understand the intricacies of production, there is a problem.

Asking The Right Questions

  • Align with Real Scenarios

Our platform addresses the entire range of eCommerce businesses, spanning from product companies selling apparel, jewellery, or an entire marketplace, to ecommerce services like salon listings, travel, real estate, hotels and others.

In such an environment you are looking to earn specificity with a generic outlook. So, while a generic solution will give average results, the platform needs to have multiple layers underneath, to handle specific situations, addressing the need of intelligence.

The interviews we conduct test if the candidate can deal well with ambiguity, see the bigger picture, divide it into logical smaller portions and optimize for best results. You don’t want someone who gets stuck in analysis paralysis.

The best interview format that has worked for us has been to ask open ended questions. We have asked the same set of questions to over 200 candidates. There are at least 20 different ways to start approaching the answer.

The idea is to ask a question which opens at least 10 different layers of complexity beneath it. This allows the candidate to drive the interview and choose their direction.

  • Bar-raiser Rounds

Machine Learning is a very dynamic space. With new developments happening every few days, it is necessary to find people who can adapt and unlearn. We typically have 5 rounds of interviews of which the last round is a bar raiser. The objective is to ask questions which defy the candidate’s expectation and test how they react when the situation goes out of hand..

  • Startup Culture Fit

Even after they have passed all the technical rounds, we have declined a few candidates because we weren’t convinced about the culture fit. When you are a startup, you want people passionate about building things from zero. The hunger to learn, adapt, create, and slog becomes just as important. That passion, in each person who joins us, stands to be of paramount importance.

In the end, hiring is the make or break for a startup. A great team will build an amazing product and customers will love it. It is hard,  but is worth the effort.

Note: The views and opinions expressed are solely those of the author and does not necessarily reflect the views held by Inc42, its creators or employees. Inc42 is not responsible for the accuracy of any of the information supplied by guest bloggers.