Creating a successful blueprint for fundraising involves a meticulous and strategic approach.
At TechSparks 2023, the destination for deliberating, deep-diving, and understanding the promise of India’s Great Indian Techade, a panel of eminent experts shared how we can collectively shape a future empowered by AI. The panel included Ravi Chhabria, Managing Director, NetApp India; Anurag Seth, Principal AI/ML Advisor AWS Startups; and Sangram Sabat, COO and Co-Founder, Saarthi.ai.
Unleashing the power of data in AI and ML startups
Chhabria kickstarted the discussion by highlighting the impact and focus of the small but impactful NetApp Excellerator, which has supported around 80 startups, with approximately eight of them achieving significant success in terms of exits or investment rounds.
“The primary focus lies in data at scale, providing innovative companies with the tools and support to make a positive impact in the world. Notable startups include Neuro Saffer, Brain Sight, and Live In Sense, each dedicated to advancing data-driven solutions in their respective fields, such as neurosurgery, brain activity mapping, and industrial automation,” Chhabria said.
Data plays a pivotal role in the realm of AI and ML startups, and acquiring quality data is a fundamental challenge. Seth explained that many startups initially rely on open-source data and models available on platforms like Kaggle and DataDog as starting points. However, the transition to production may necessitate generating synthetic data through algorithmic methods. Additionally, startups often begin data collection both internally and from third-party sources, continuously enriching their data assets.
<div class="externalHtml embed" contenteditable="false" data-val="
“>
A noteworthy advantage is that generative AI applications are relatively easier to build, as few-shot learning properties in large language models allow for fine-tuning with a limited amount of data, making them efficient and practical.
An interesting case study involves anomaly detection for surveillance cameras, where the startup initially used statistical algorithms to learn normality without pre-trained models or tagged data. As anomalies occurred, they collected data, creating their dataset.
“They then adopted pre-trained models for scene understanding and augmented them with collected data to accelerate the path to production while continually improving their models. These approaches allow startups to kickstart their AI and ML ventures effectively while simultaneously building a valuable data mode,” Seth shared.
Navigating complex linguistic challenges
Today, there’s a notable focus on the Indic language ecosystem and understanding how Indian speakers communicate with machines. Sabat has been a pioneering force in the domain of Indic languages for over six years. Saarthi.ai has successfully addressed the unique challenges posed by India’s linguistic diversity, obtaining and managing extensive data and resources.
“When solving the human-machine communication challenge, being human-centric in design is crucial, especially in a linguistically diverse country like India, where heavy accents, domain-specific terms, and compound words present unique complexities. Collaborating with global researchers and bringing their expertise to Indic languages has been our key to success, enabling us to serve major financial institutions and fintechs,” he said.
He said building a technology lab is one thing; taking it to an enterprise scale with privacy, security, and consistency is another. “We’ve made a significant impact in debt collection, but there’s much more to achieve in India’s vast market.”
Powering India’s AI revolution
India’s future in AI is promising with a wide range of applications across sectors like smart transportation, airport security, and autonomous vehicles. Challenges exist, especially related to data, but Indian startups are diligently working to solve these issues. The future involves leveraging India’s AI talent and converting it into an AI natural resource.
Ensuring the responsible development and deployment of AI applications requires diligent efforts across all phases, from design and development to deployment and monitoring. Legal and ethical considerations are paramount, and educating users about the potential risks is essential.
The key is to remain grounded in addressing grassroots problems while developing AI responsibly and ethically.
