In the realm of AI startups, India and South Korea showcase contrasting landscapes. With Asian countries emerging as a strong hub for AI, South Korean AI startups are observing a wave of funding, propelling the industry to new heights.
Rebellions, a South Korean fabless AI chip startup, captured significant attention by raising $124 million in a Series B round, bringing their total funding to an impressive $210 million.
A few reports suggest that South Korea is fast emerging as a powerhouse with 1,101 AI startups so far in 2024, which include KRAFTON AI, Mathpresso, Zig Zag, Lunit, and Channel Talk.
Along similar lines, AIM Research suggests that as of 2024, India was home to over 100 GenAI startups. While on the whole, there are around 750+ AI startups in the country.
The funding story portrays a scene of diverging fortunes.
Indian AI startups, encompassing infrastructure and services, dropped nearly 80% in 2023 to $113.4 million from $554.7 million in 2022, according to a Tracxn data.
In contrast, South Korean counterparts witnessed a resurgence, raising 326.8 billion won ($240 million) collectively in the first quarter of 2024—up from 89.8 billion a year earlier, as per The VC, a startup investment tracker.
Despite these financial fluctuations, the race for the unicorn status reveals a clear lead. While no South Korean AI startup has yet attained unicorn status, Krutrim from India achieved this milestone in the first quarter of 2024.
One South Korean AI startup that stands out amid all these remarkable developments is Upstage. With a fresh $72 million in funding in the first quarter of this year alone, Upstage achieved new orders worth about 10 billion won for its Document AI solution and Solar LLM API, doubling the orders it received in 2023.
Their unique business model boasts over 400 customers, and their recent expansion to the US has introduced domain-specific LLMs, particularly specialised language models (SLMs).
Upstage to Upscale
Founded in 2020, Upstage began as a document heritage provider, collaborating with major insurance, banking, and logistics companies to transform Korean documents and PDFs into machine-readable data.
Existing models like GPT struggle with Korean language data, which influenced Upstage to develop their own LLM, Solar.
This model was crafted using several innovative techniques, including “upscaling” (US), allowing them to patch different model sizes together and train with a local Korean language corpus.
The scarcity of high-quality Korean data, comprising less than 0.5% of web data, presented a challenge, but Upstage overcame it by building a sophisticated, culturally aware model tailored for enterprise use.
“Creating such an advanced model required more than just cutting-edge modelling techniques; it necessitated a robust ecosystem. This ecosystem included data providers, cloud services like AWS, orchestration tools, strategic partners, and system integrators,” said Casey Jones, founder, CJ&CO, at the AWS event.
In July 2023, Upstage achieved the top spot in the global evaluation of large language models. Their 30-billion-parameter LLM scored 67 points, surpassing Meta’s LLaMa-2, which scored 66.8, on Hugging Face’s Open LLM Leaderboard.
Despite having fewer parameters, Upstage’s model outperformed those from big tech firms like Microsoft, Stability AI, and Databricks by about 10%. Additionally, Upstage’s model scored 56.5 points for mitigating AI hallucinations, exceeding Meta’s 52.8 points.
What Can India Learn?
While attending the Machine Learning Developers Summit (MLDS) organised by AIM, GiJung Kim, the CEO and founder of Align AI, observed a vibrant community of Indian engineers and entrepreneurs actively engaging with frontier technologies, including AI chatbots.
“We began forging connections with Indian entrepreneurs. They were able to find our product and they started signing up,” Kim mentioned, highlighting the organically piqued interest in Align AI within the Indian tech community.
So, a major concern for Indian AI startups is the lack of understanding of the customer market. Many AI startups are unsure about the long-term direction of their products or services. Without this clarity, it’s challenging for venture capitalists to be convinced to provide funding.
India has made strides in AI research and innovation. Between 2010 and 2020, the country filed around 5,400 AI patents across various techniques, with a notable surge—over 94% of these patents were filed in the last five years, indicating accelerated activity in AI innovation.
However, the development of AI solutions in India is hindered by a shortage of skilled AI researchers.
According to the Observer Research Foundation, the country faces a shortfall of AI professionals, estimated at 213,000, even though it ranks second globally in producing master’s-level engineering students, surpassing the US in numbers.
But things are slowly changing. As of 2022, Indian software developers contributed significantly, accounting for 24.19% of AI project contributions on GitHub, outpacing contributions from the EU, the UK, the US, and China.
