Forty-five minutes of hold music is a modest contribution to the history of artificial intelligence. For Umakant Soni, it was enough to help start a company. He had called a bank with a simple question and spent the better part of an hour waiting. The experience suggested a business opportunity: perhaps software could manage a customer conversation with more attention than the existing arrangements seemed to permit.
In 2009, he co-founded Vimagino. Its software agents were called “angels”, a forgiving name for something expected to deal with online shoppers. They were designed to pick up emotional feedback and improve sales and support conversations. Long before the current enthusiasm for conversational AI, Soni was working on the old, stubborn problem of getting an answer when you needed one.
His ambitions have since acquired considerably more floor space. He co-founded pi Ventures, became the founding CEO of ARTPARK, and is now a co-founder of Bharat1.ai, which is developing AI research infrastructure and proposing an AI city in Bengaluru. The progression has an underlying question: how much must you build around an invention before it can do useful work?
The helmet before the chatbot
There was an earlier prototype. For his MTech project at IIT Kanpur in 2001, Soni developed a pollution-free helmet. He travelled to Bengaluru to pitch a business plan to Infosys co-founder N. R. Narayana Murthy. The episode puts his later career in perspective. Even as a student, he was prepared to carry a technical project into a business conversation, where an elegant idea would have to answer less elegant questions.
At Wipro, he worked on customer-service technologies and helped incubate businesses within the company. He left in 2008. Starting independently meant giving up an established organisation’s resources while taking responsibility for the expensive interval between having an idea and finding a market.
Vimagino’s colleagues included fellow IIT Kanpur graduates with experience at Adobe, Google and Yahoo. Their previous employers supplied an impressive introduction. They could not supply the future. The team had to persuade customers to try unfamiliar software and investors to finance a product company. A résumé can open a door; it has limited power over the purchase order.
The angels needed an ecosystem
One early Vimagino pilot served Wipro’s internal help desk. That was a useful setting for discovering the gulf between understanding words and understanding a request. Software dealing with a Microsoft-related facilities complaint had to distinguish it from a request for training in Microsoft technology. Everyday language had brought its usual assortment of traps to the demonstration.
Former Wipro executive Mythily Ramesh became a mentor and angel investor. Her support supplied something a capable technical team could not manufacture by writing more code: an experienced backer willing to help a young company through its uncertain early stage.
Looking back in 2022, Soni was blunt: “we could not sell the technology.” He described spending a year in Silicon Valley studying how technology companies emerged there. His explanation moved beyond the individual product to the surroundings that let a product become a company. Funding, customers, research and people with relevant experience had to meet somewhere.
“we could not sell the technology.”
Umakant Soni, reflecting on his early chatbot venture, 2022
An investor’s seat at the table
In 2016, Soni and Manish Singhal founded pi Ventures. Both were IIT Kanpur alumni; Singhal had co-founded LetsVenture. They chose a focused investment thesis around artificial intelligence, machine learning and the Internet of Things. Soni had moved to a position from which he could help finance other builders.
The fund’s early thinking was anchored in applications. It looked for a business use case, a defensible technical approach and a coherent account of how data would be used. A company needed to explain what its intelligence actually accomplished. That requirement sounds obvious until one considers how comfortably the phrase “powered by AI” can occupy the space where an explanation ought to be.
The International Finance Corporation’s 2017 investment disclosure identifies Soni and Singhal as the founders of the investment manager. Its description placed enterprise automation, logistics, retail, smart cities and other applications within the proposed remit. AI was being treated as technology that could travel across markets, with the investment vehicle supplying an organised way to back it.
Build the product
Finance the builders
Connect research to use
Build a testing environment
A place around the invention
ARTPARK widened the scope again. Established in association with the Indian Institute of Science, the AI & Robotics Technology Park brought researchers, entrepreneurs, industry and government into the same institutional effort. Soni was a co-founder and its founding CEO. The task included helping research make the difficult journey into applications people could use.
For a robotic system, that journey is especially demanding. A prototype must work somewhere less accommodating than the place where it was assembled. It needs components, practical experience, repeated testing and someone willing to use it. The work between a promising result and a dependable product can be lengthy, expensive and poorly suited to applause.
ARTPARK’s current work spans robotics, industrial systems, language technology, communications infrastructure and an academy. Its incubation programme describes support across different stages of a startup’s development. That breadth gives the institution a practical purpose: a researcher looking beyond a laboratory needs access to several kinds of help, often in an inconvenient order.
Soni also worked across national boundaries. The Sweden-India Business Council appointed him Senior AI Advisor, and his public biography records India’s nomination of him to the Global Partnership on AI’s Innovation and Commercialisation working group. These roles fit a career increasingly concerned with the arrangements surrounding technology: who supports it, how it travels and what lets it enter wider use.
A robot wanders into the argument
In 2024, a family conversation offered him another way to frame the problem. His daughter had watched The Wild Robot and described it as something like The Jungle Book, with a robot in Mowgli’s place. The comparison led Soni to consider how an artificial system learns to live among others, and what happens when it enters an environment nobody has prepared for it.
He proposed an AI city: a bounded setting where autonomous systems could operate under observation, with infrastructure and rules designed for experimentation. Mobility, logistics, education and city services could provide practical settings. The point was to learn from interactions while retaining ways to limit an experiment if it went wrong.
It is a revealing extension of the original customer-service concern. A chatbot encounters a confused question. A robot encounters a shared physical world. Both must operate among people whose behaviour resists tidy assumptions. Soni’s proposed answer now includes the environment in which the machine is expected to learn.
The experiment gets an address
By 2026, that proposal had an organisational home in Bharat1.ai. Soni co-founded the company with Subhashis Banerjee and Sireesh Kupendra. Its B1 AI SuperPark is based at RGA Tech Park in Bengaluru. The company presents its work as an effort to bring researchers, startups, enterprises and universities together around AI infrastructure and experimentation.
The larger plan includes a proposed 70-acre AI city. Public presentations have described a living laboratory with people, AI agents and robots generating physical-world data for research. Those are plans and targets. They describe the experiment Soni wants to construct, with the results still to be established.
At the Entrepreneur India Tech & Innovation Summit in 2026, he discussed the possibility of a research setting comparable in spirit to CERN. The useful part of the analogy is the shared setting: specialists and equipment assembled around questions that would be difficult to pursue alone. Whether an AI campus can produce that kind of collaboration will depend on the work its participants actually do.
A diagram of the ambition behind the AI city, rather than a claim of completed results.
What will people do with the time?
Soni’s argument for an Indian AI approach starts with people. In his 2026 public remarks, he emphasised enhancing human capability and questioned the wisdom of copying approaches developed for different economies. His proposed ecosystem puts human participation near the centre of the design, alongside agents and physical machines.
He has also argued for distributed intelligence and systems that reflect India’s linguistic and cultural variety. In an April 2026 interview, he warned that India had a limited window to organise its own AI capability. The urgency concerns institutions as much as models: developers, capital and infrastructure need ways to work together.
His writing ventures into a more personal question. If machines release people from some of their work, what will fill the available time? In an October 2024 essay, he recalled reading Sri Aurobindo during a visit to Auroville roughly two decades earlier. He returned to questions of human purpose, empathy and education while considering the growing abilities of machines.
He paints abstracts in his spare time. That interest sits comfortably beside the less measurable concerns in his essays. The investor who asks what a system can produce also wonders what its users will value. An extra hour can be counted; deciding how to spend it remains a human occupation.
In September 2026, discussing whether India should slow its AI efforts, he said: “We have not even started.” It is a characteristically impatient assessment from someone who has already spent years moving between products, investment and institutions. His earliest problem involved a person waiting for an answer. His current work asks what a whole place might need before better answers become possible.