Being a former business technology analyst, Siddhant Sadangi’s career in data science was a happy coincidence. Sadangi worked on pet data science projects in his free time, which helped him snag a job as a data scientist at Reuters. Later, he went on to work as an ML developer advocate in Polish startup Neptune.ai that helps enterprises manage model metadata. Sadangi, who has also done part-time projects as a journalist, believes in his own industriousness more than anything else. 

Analytics India Magazine caught up with Sadangi to understand what it takes to be a data scientist today and what actually works in the AI/ML industry. 

AIM: As an ML developer advocate in Neptune.ai, what exactly does your role entail? 

Siddhant: A developer advocate is someone who advocates for the developers using a product. It is our job to make sure that the developers are able to use a product to its full potential and act as a bridge between the developer community and the company. The developer relations team within Neptune.ai does just that. 

We work with the developer community, and our engineering and product teams to create material that helps developers use our platform better. These can be documentation pages, blogs, code examples, or synchronous forms of knowledge exchange like live demos, onboarding, and support calls. We also work on developing integrations with other open-source tools to facilitate seamless interoperability between Neptune and some of the most popular ML libraries and frameworks.

AIM: How has your role evolved from being a data scientist in a news organisation like Reuters to working in Neptune.ai?

Siddhant: From working in a 170-year-old company with over 24,000 employees to a four-year-old start-up with a headcount of less than 50, it has been quite a change! With Reuters, most of my work was directed toward non-technical internal stakeholders. With Neptune.ai, most of it is for our external clients and users who are very technical and hands-on. Also, since Reuters is a much larger organisation, the role also involved some ancillary work which is sometimes not what you want to do (at least at this stage of your career) but is important for the organisation.

With Neptune.ai, the work is a lot more focused, and I get more ownership in choosing what and how I want to do things. Both roles have given me different perspectives and motivation to do the same thing differently depending on who the end-user would be.

AIM: What were some of the biggest challenges you faced in your career?

Siddhant: Unlike more traditional technical roles, since there is almost no barrier to entry for a data scientist – most employers expect you to have at least some hands-on experience. Just prepping with sample interview questions is not enough to get in. I was lucky that during my time with Deloitte as an analytics consultant, I had opportunities to work on data science projects. This gave me confidence during interviews and helped me tackle questions around real-world scenarios.

AIM: What are some of the most important lessons you have learned in your career that you can give aspirants in the field? 

Siddhant: First, as mentioned earlier, just theoretical knowledge of the concepts won’t help you get into a good company, especially with the kind of supply of labour we have these days.

Also, perhaps the most important lesson once you have finally broken in, is to understand that not every problem is an AI/ML problem. In the age of GPT and DALL-E, most of the business value is still generated by traditional ML and statistical approaches like regression. In my over four years as a data scientist, I have never used a deep learning solution to solve a business requirement. Sometimes the best solution is also the simplest. Don’t chase the latest and shiniest algorithm. Develop your soft skills, talk to stakeholders, and understand the requirements. 

AIM: What is it about journalism and data science that draws you in?

Siddhant: My interest in data science is actually independent of my interest in journalism. I had been passionate about environmentalism right from the school days. That gave me an opportunity to work with international publications as a journalist and interview positive changemakers from across the globe. When I was introduced to model United Nations in my college days, I was naturally attracted to the Press Corps there, and it was then that I first learnt about Reuters.

Coming to data science, I also always had an affinity for coding. I took up a few data science courses online during my early-career days and liked the opportunities it offered. So when I found the opening for a data scientist at Reuters, it felt like the perfect fit for me.

AIM: What are some of the biggest trends across AI/ML currently? 

Siddhant: The beauty of this field is how fast trends evolve, especially on the consumer side. In 2020, text-to-text was in vogue due to GPT-3. In 2021, DALL-E 2 brought text-to-image to the forefront. Stable Diffusion and Midjourney reinforced this trend this year, only to see Google’s Imagen and Meta’s Make-A-Video divert the limelight towards text-to-video. All this in just a span of two years! 

On the organisational side, as the field matures, we see people talking a lot about MLOps, and this is reflected in the number of MLOps-related tools entering the market. Finally, on the policy front, AI Ethics is being discussed a lot. When you want AI to help make policy decisions, you do not want it to have the same biases which humans inherently do, and on which data the AI is trained.