Ankit Maheshwari, who serves as Chief Product and Technology Officer and is a member of the founding team at Innovaccer, discussed how his interest in artificial intelligence (AI) began at a moment familiar to many tech professionals: when code first appeared to him as a source of power.
In school, when he was around 10th grade and was introduced to the basics of programming, computers ceased to be merely a subject of curiosity and transformed into an opportunity. He recalls: 'It was the first glimpse of how just a few lines of code could create something real.' This experience shaped all subsequent decisions, including his choice of subjects and career path in technology.
By his second year in engineering college, when AI and machine learning (ML) in India were largely academic concepts, Maheshwari was already moving beyond theory. His college's software incubator became a testing ground. He remembers that they operated like a 'mini-startup on campus,' creating solutions not only for their educational institution but also for neighboring institutes. Together with his peers, he experimented with early intelligent systems—from a rudimentary tool resembling GPT that transcribed and indexed lecture notes, to a camera-based attendance system operating on rule logic and basic machine learning.
These early projects not only honed his technical skills but also solidified Maheshwari's conviction, which later defined his work at Innovaccer: intelligent systems are meant to augment human capabilities, not replace people.
Innovaccer was initially a company focused on horizontal data and analytics; its name derives from the phrase 'innovation accelerated.' In its early stages, the startup collaborated with academic institutions such as Stanford University, Harvard University, Wharton School, and the Massachusetts Institute of Technology, helping researchers test hypotheses using clean, analysis-ready datasets. A key realization was that even the most advanced machine learning models depend on the quality of the data they are built upon.
Working across various sectors, including projects with NASA and Walt Disney, the team encountered its first healthcare client in Des Moines, Iowa. This interaction revealed a deeper problem: despite the US spending nearly $4 trillion annually on healthcare, data is highly fragmented, limiting providers' access to patient medical histories and real-time information during care delivery. This resulted in inefficiencies and excess costs amounting to $1.2 million.
Recognizing the scale and human impact of the problem, Innovaccer decided to focus exclusively on healthcare. The company set a goal to create intelligent systems that unify disparate data, provide predictive capabilities, and support better clinical decision-making—a mission it has pursued for ten years.
As Chief Product and Technology Officer at Innovaccer, Maheshwari oversees product strategy and technological architecture, adhering to a clear philosophy: AI must be a fundamental element, not an add-on layer. Instead of bolting chatbots onto legacy systems, Innovaccer develops applications built from the ground up on AI, starting with data unification, analytics, long-term patient context, and finally, actionable insights for service providers.
Maheshwari asserts: 'AI is only as good as the data and infrastructure behind it. Clean, connected data is the real differentiator. And that is our secret ingredient: our systems were not created overnight.' These long-term investments now serve seven out of the ten largest healthcare systems in the US, covering over 80 million lives on the Innovaccer platform.
For Maheshwari, the biggest hurdle to widespread AI adoption in healthcare remains the data itself. Systems built decades ago on outdated infrastructure continue to operate in silos, hindering access to clean and interoperable data. Without unified and reliable datasets, even the most advanced AI models risk producing misleading results.
Regulation adds another layer of complexity. Medical organizations operate within strict regulatory frameworks and approach new technology adoption cautiously. This risk aversion often slows down experimentation and large-scale implementation.
However, perhaps the most challenging task is change management. Clinicians operate under intense pressure and rely on workflows refined over years. Implementing new tools requires more than just technical competence; it demands trust, demonstrable value, and seamless integration. Maheshwari notes: 'You can build something powerful, but asking clinicians to change their daily workflow is a tough sell if it doesn't fit naturally into their routine.'
Maheshwari believes that AI is at a point of transformative inflection, similar to the industrial and internet revolutions. In the short term, it may cause disruptions, from workforce changes to energy consumption issues. But over time, he expects AI to help solve many of these problems, including breakthroughs in renewable energy and efficiency.
In healthcare and life sciences, the potential impact could be enormous. AI can accelerate drug discovery, advance cancer research, and extend healthy aging, reducing timelines that once took decades to mere years. Beyond healthcare, its application in agriculture, climate change resilience, and education could transform societal approaches to food security and resource scarcity.
Maheshwari is particularly interested in two areas. The first is edge AI, moving intelligence from energy-intensive cloud systems to efficient, local models that are faster, safer, and more scalable. The second is the transition from digital AI to physical AI through robotics. He points to initiatives like Tesla's Optimus as evidence of embodied intelligence capable of functioning in the real world. For Maheshwari, these advancements mark the next leap in how AI will transform industries and human potential.
Amazon Web Services has been a foundational partner in Innovaccer's journey since its earliest days. As the company deepened its focus on AI, platforms like Amazon Bedrock and Amazon SageMaker allowed for rapid model experimentation without being tied to a single ecosystem. This flexibility aligns with Innovaccer's technologically neutral approach, enabling engineers to test and implement what works best for specific healthcare use cases.
Speed is also critical. Scalable cloud infrastructure allows solutions to move quickly from prototype to production, ensuring teams spend less time managing backend complexity and more time solving customer problems. Maheshwari emphasizes: 'We want to be the fastest innovators, delivering solutions to customers quickly. If we get stuck trying to figure out how to scale, we lose focus on the problems that matter.'
AWS also provides enterprise-level capabilities in terms of security and compliance, reducing operational overhead while meeting the stringent requirements of healthcare systems. This enables Innovaccer to innovate with confidence through faster experimentation, seamless scalability, and trusted compliance.
To keep pace with the rapid evolution of AI, Maheshwari uses AI itself. He has created a set of intelligent agents that scan global events every morning, tracking advancements in models, agent frameworks, and healthcare innovations, and summarizing them into brief reports. When something catches his attention, he delves deeper, either by requesting further research or investigating it personally.
Outside of work, Maheshwari is an avid cricket fan and enjoys reading non-fiction, especially biographies and autobiographies of leaders, which he takes on long flights. Away from technology, he finds poker particularly engaging—a game that combines probability, strategy, and human psychology.
ByteDance is developing a new artificial intelligence model that could reach ten trillion parameters, approaching the most advanced technology associated with Anthropic, according to exclusive reports from the Financial Times.
This advancement is taking place in China, set against the backdrop of intensifying global competition for more powerful AI systems. If confirmed, the model from the owner of TikTok would exceed more than three times the size announced for Kimi K3, developed by the Chinese startup Moonshot AI, which has 2.8 trillion parameters.
This initiative reflects the strategy of Chinese companies aiming to increase their competitiveness against US rivals, seeking to create larger models without incurring excessive operational costs.
ByteDance has already begun the pre-training phase of this model, a process that generally requires three to six months before subsequent refinement stages and possible public release.
The number of parameters serves as a metric to evaluate the magnitude of an AI model; these components represent mathematical configurations learned by the system from data, helping the technology identify patterns, generate responses, and execute tasks, although the isolated quantity of parameters does not define the tool's total capability.
The company's progress follows a series of investments made by Chinese companies in the field of large AI model development. Before the launch of Kimi K3, other Chinese models, such as Meituan's LongCat-2.0 and DeepSeek V4-Pro, were among the largest in the Chinese sector, both having 1.6 trillion parameters.
Direct comparisons between Chinese and American models remain limited, as corporations like Anthropic and OpenAI do not disclose the parameter count of systems such as Mythos, Fable, or GPT-5.5.
Industry estimates cited by the Financial Times suggest that Anthropic's Mythos 5 has approximately eight trillion parameters, while Fable 5 has about five trillion. Under this scenario, the ByteDance project would be close to the scale attributed to the American company's most sophisticated system.
