In 2017, Niki Parmar was part of research at Google that introduced the basis on which ChatGPT and other AI systems are built. Now she is part of companies that build AI tools and products
Niki Parmar, Co-founder, Essential AI; co-founder Adept AI Labs; Technical Staff, Anthropic
Image: Courtesy Essential AI
In 2017, a research paper titled ‘Attention Is All You Need’, co-authored by Niki Parmar along with seven researchers at Google, transformed the way in which artificial intelligence (AI) was being looked at and used by the world. The paper introduced a revolutionary neural network mode called ‘Transformer Architecture’, which became the basis for many modern AI systems, including large language models (LLMs) like ChatGPT.
Parmar, who had joined Google in 2015, was the youngest member of her team and the only one without a PhD, according to an NDTV report. During the interview with NDTV last October, Parmar, who worked at Google for almost seven years as an engineer and research scientist, spoke about the intimidating nature of the atmosphere in Google during her initial days, and also about it being an incredible learning experience. “I was constantly surrounded by people pushing the boundaries of what AI could do,” says Parmar, who believes one of AI’s core purposes is to explore technology’s untapped potential.
The learnings at Google have held her in good stead. Today Parmar is a well-known name in the field of AI, working to strengthen its growth and development. A member of the technical staff at Anthropic, a US-based AI company, she is also a co-founder of a couple of other AI startups in the US—Essential AI and Adept AI Labs.
Anthropic, where Parmar provides her expertise in building reliable and interpretable AI systems, conducts researches in areas such as natural language, human feedback, scaling laws, reinforcement learning, code generation, and interpretability. Adept AI Labs, launched in 2022, where Parmar is also the CTO, is an enterprise AI tool that enhances productivity by managing manual, repetitive workflows across the tools that corporate organisations use daily.
Her most recent venture Essential AI (2023) is working on developing full-stack AI products that quickly learn to increase productivity by automating time-consuming and monotonous workflows related to data processing. The company is backed by a number of tech giants including Google and Nvidia.
Parmar, who was always interested in computer science and AI, wanted to get into an IIT, but when that didn’t happen, she did her Bachelor of Engineering in information technology from the Pune Institute of Computer Technology. It was also in Pune that her passion for machine learning came to the fore as she became more and more fascinated by the potential of data, pattern recognition, and optimisation.
She taught herself the basics, and applied for a master’s degree at the University of Southern California (USC) to pursue a course in computer science. There she explored how machine learning and big data could be applied to social science questions, further deepening her interest. Soon after graduating from USC, she joined Google where she worked on end-to-end deep learning systems, focusing on alternative approaches to natural language processing using transferable embeddings, task-specific optimisation, and weakly supervised learning.
“As I progressed through these experiences, deep learning models consistently emerged as powerful tools for a wide range of challenges,” she told Medium in an interview in May 2020. These insights ultimately motivated her to transition into pure research, where she continues to expand her knowledge and make meaningful contributions to the field of AI.
Though most of her work is in the US, Parmar is also keen to contribute to the growth and potential of AI in India. During a chance meeting and conversation with Aakrit Vaish, founder of India-based AI company Haptik, at the Nvidia AI Summit nine months ago, Parmar expressed interest in contributing towards the AI landscape in India. Vaish, who was an advisor for the IndiaAI Mission at the summit, has remained in touch with Parmar since then, and has had brainstorming sessions with her around the potential of AI in India.
He says Parmar’s work in AI so far has had a wide ranging impact on the field globally.
“There would be no ChatGPT without the T in it, which stands for ‘transformers’, which is the research Niki was a part of. It should be a matter of great pride for the country that a person of Indian origin was responsible for such pioneering work,” he says.
According to Parmar, AI is far from reaching its limits and is advancing at an exponential rate. In an interview with Moneycontrol during the Mumbai Tech Week this February, Parmar spoke about new capabilities emerging within AI, such as the computer user guide and concept, where models learn to interact with computers, but also pointed out that access to high-performance computing resources remains a key challenge in AI development.