HT interview: ‘We have enough time to experience, adapt to AI,’ says AI researcher
Today Parmar is the reigning queen of AI, one of a handful of women founders in the ultra-alpha, ultra-competitive global AI space
Every AI model of note – ChatGPT, Bard, Krutrim – is built using the Transformer model. (The T in ChatGPT, stands for Transformer). This founding model was thought of in 2017 by eight AI researchers at Google over lunch, and published in an iconic open-source scientific paper titled “Attention Is All You Need”. It was this system that made it possible for machines to generate human-like text and images – or become artificially intelligent.

Niki Parmar, who was born in Pune, and moved to the US in 2013 for her Master’s at the University of Southern California, was the youngest developer of the Transformer model and the only woman in that cohort of researchers.
Today Parmar is the reigning queen of AI, one of a handful of women founders in the ultra-alpha, ultra-competitive global AI space. A CTO and co-founder, her second startup, Essential AI, raised $56.5 million in Series A funding three months ago.
Parmar sat down for an interview with HT in the noisy but cosy Farley’s Cafe in San Francisco’s trendy Potrero Hill Valley on a rainy Saturday afternoon in February. The 33-year-old was dressed in a brown sweater under an oversized blue parka and running shoes -- looking the quintessential Silicon Valley geek in translucent plastic prescription glasses with her hair tied up in a casual ponytail.
Edited excerpts:
At 18, as the only child of middle-class parents in a family where women get married early, you already knew you wanted to be in mathematics and do problem-solving. But there was no one in your family who was an engineer, so what made you so self-driven?We had limited means growing up, but my father never curbed my ambition. He made sure I could get access to the right education and resources, even by taking on debt. He always encouraged me.
You appeared for JEE (Advanced) but didn’t get into IIT. How did you feel at that stage?That my life was over (laughs). I was completely confused. Finally, I decided not to take a gap year and instead joined the Pune Institute of Computer Technology to do a bachelors in engineering and focus on doing side projects to learn more. The cloud revolution was just beginning in 2010 and that’s when I took a Udacity course, “Introduction to Artificial Intelligence” by Stanford professor Sebastian Thrun and Google’s Peter Norvig. That was my first foray into machine learning.
Machine learning hooked you, didn’t it? A couple of years after completing your engineering, you headed to California to pursue your Master’s in Computer Science and continued to work on natural language processing (NLP). You joined Google Research in the Bay Area as you wanted to pursue machine learning (ML). How did you start working on the Transformers model? At Google, I joined a team of PhDs to build models that look for two queries similar to each other – a concept called semantic similarity. One day, my manager offered me a project where the team had been working on discovering new ways of doing sequence-to-sequence models to work on automatic language translations. I was like, yeah, that sounds interesting and jumped on it. This became the Transformers model – a new category of models for artificial intelligence. It changed the game.
You released the Transformers model in 2017. At that time, Google was at the top of its AI game but in 2022, OpenAI used the same model and brought out a viable product in ChatGPT. In a way it swept the carpet under all of you. How did you feel about that? Open AI had the vision to figure out how the Transformers model could do certain tasks if you trained it at that scale of data. However, ChatGPT also showed how much is possible with the model we developed. I have gained a lot from being recognised as a researcher who worked on the Transformers model.
Let’s talk about Essential AI, your second startup. You began it in 2022 with Ashish Vaswani, the primary author of the Transformers paper. You have raised more than $60 million in two rounds of funding and have announced a partnership with Google Cloud a month ago. How did you celebrate? We didn’t! The very next day that we got the funding, we took out our laptops and started to work to build a product for our startup. We have enough funding to build our model and hopefully later this year we can have a prototype. That’s when we will celebrate.
There are not many women AI researchers and even fewer AI founders. You were the youngest in the Transformers model team and the only woman. And now you’re a startup founder in the technology industry which is celebrated for its toxic masculinity and gender bias. How do you deal with that? I have learnt to face some issues and fight them out, which always requires energy, but also ignore some issues. I want to work on problem solving, to develop my AI product and not be a token in a panel because of my gender. It’s a delicate balance.
It must be hard being a young AI founder in a cutting-edge, aggressive and very, very competitive industry, even by Silicon Valley standards. How has that affected you? There’s a lot of pressure. People look at everything you’re saying, doing, what you know and don’t know, where you’re confident or not. It’s hard to constantly give this energy, show clarity as a leader to your ambitious employees as there are so many ambiguities. You’re figuring out so many things. The journey can be very isolating but you also discover a lot about who you are, what you care and what you’re capable of. Being a founder has given me more confidence, more belief in my own capabilities.
Like any immigrant you’re living two lives – one in the Valley and one in India. I’m sure you miss India and your family, and then you miss the work and life opportunities that Silicon Valley present to you in your field. Do you ever think of coming back to India?I came to the US to be financially stable and build something for myself. Now I’m in the next phase of my life, where I’m building a startup. I love the Bay Area because there’s a lot of intellectual stimulus for me here. However, I spent a majority of my life in India and most of my family is there. I don’t have a plan about this, but as the next phase in my life, I want contribute to India’s growth, provide more opportunities for people to learn, and encourage the amazing talent there. Especially girls.
What’s the future of AI like? The good, bad, and ugly... We will get much more easy access to knowledge and skills than we have today. If I want to do something that I’m not an expert at, AI is going to close the gap for me. Through AI we will elevate ourselves, achieve more, learn more and do more.
Will we lose our jobs?Over the long run, AI will automate certain things and certain jobs will cease to exist. Physical jobs will remain however for a lot of cognitive work, we will be forced to specialise, to become an expert in something That happens every time there’s a new technology. Like always, people who adapt will survive, and people who don’t might get uprooted.
But isn’t technology forcing people to adapt faster and faster than they can?We feel like the technology’s coming faster, but it never does. AI will take five to 10 years to be integrated into our work and lives. We have enough time to experience it, adapt to it, and change.
Shweta Taneja is an author and journalist based in the Bay Area.
