Choosing Machine Learning Career Path

This article talks about the job opportunities that are available, how ML is evolving and what kind of companies are best suited.

Roles Available:

• Data Scientist Role: The role of data scientist is to get the data, try to experiment with different ml approach and come up with a better model for the problem. Their work is getting delieved as insights for business decision or model is consumed into the product
• Machine Learning Engineer: The role of Machine learning engineer is get the best model from data scientist. Create api service and ml pipeline that optimize for throughput and latency.
• Data Engineer Role: The role of data engineer is to build the data pipeline and data from this pipeline are used by data scientist to build the model

Career Question

Technology is changing fast. Day to Day activities of Data scientist and ML Engineer roles are getting automated or erased. What should one do?

1. Create a niche where the generalist ML tools could not automate you. Say you become Data scientist in Healthcare or Finance
2. Barrier for software engineer to work on data science is decreasing. Machine Learning engineer are definitely going to learn data science. Data scientist have to learn software engineering

How useful is the Kaggle for the Career?

1. Kaggle badges are clearly differentiates you among others when you have few years of experiences.
2. After a few years in the ML space, your end to end understanding of the system i.e data extraction, model monitoring etc is important

What level of the mathematical concept I know?

if you are in product development (not R&D), having good intuition is enough. I personally would recommend to understand how to tweet different tools. For example, huggingface takes text as input, how will I add additional numerical features (no of followers) to hugging sentiment analysis.

Choosing the Companies

Does the company has enough data or has potential to get the data?

• Many startups or company start great. They end up not doing well as the they were not able to get required data for building product

Do they have enough usecase for the ML

• They might have one or two ML cases. After that, there is no ML. It involves only ML ops and other development work. I strongly believe that Data scientist/ML Engineer requires all the jobs from building ETL to deploy the API. Yet, if there is no use case after one ML case. We have to settle for another work or look after another company

Is there push or interest in Organization to release ML Features?

• If the company sees the ML features as ‘good to have’ feature, not “must have” feature, then it is difficult to see the effort in production.