Sayantan Sarkar

In the last decade, the evolution of AI algorithms has significantly enhanced their capability to analyze medical images with remarkable precision. This progress has been particularly noteworthy in the detection and classification of abnormalities within various imaging modalities, including X-rays, CT scans, and MRIs. Despite these advancements, the landscape of AI solutions in the medical imaging market remains fragmented, marked by the existence of over 700 FDA-approved AI solutions. The FDA’s approval process is stringent, focusing on specific findings in particular body parts for distinct types of images such as X-rays, PET scans, or MRIs. The expectation is that this number will see a substantial increase in the coming years.

Introducing an AI solution into a Healthcare Provider’s (HCP) environment poses several challenges. Typically, when an AI solution is deployed, it must seamlessly integrate with the imaging hardware’s software to pre-process images and transmit them back to the hospital system. However, HCPs grapple with multiple hurdles:

1. Each hospital may possess a variety of imaging hardware, each equipped with different software systems.
2. HCPs need to conduct rigorous evaluation tests on patient data before deploying any AI solution to ensure its efficacy.
3. The integration time and associated costs with each AI solution provider are often prohibitive, discouraging the exploration of multiple solutions.
4. Hospitals often contend with limited IT bandwidth spread across various departments, making efficient integration a challenge.

Carpl‘s platform strategically addresses these challenges, offering a comprehensive solution for HCPs to evaluate, procure, deploy, and monitor AI solutions in radiology. Through a one-time integration with the hospital’s software system and imaging software (PACS), Carpl eliminates the need for multiple integrations with various AI providers. This not only significantly reduces costs but also streamlines the integration process, saving valuable time for HCPs.

Carpl’s platform functions as an AI co-pilot for radiologists, seamlessly integrating into their existing workflow. By doing so, Carpl increases radiologists’ productivity and also improves medical outcomes. The platform prioritises scans, ensuring critical patients receive timely attention, potentially saving lives.

The increasing number of imaging scans, totalling 3.6 billion X-rays globally with 2 billion in the US alone, has led to a substantial shortage of radiologists. With the growth rate of imaging scans surging over the last decade, the demand for radiologists has also increased. We envision AI playing a pivotal role in bridging this demand gap. The imaging services market in the US, valued at $120 billion with a CAGR of 4.2%, is undergoing transformative shifts. We anticipate that a significant portion of these costs will transition to AI imaging, reflecting the industry’s inclination toward adopting innovative solutions.

At the forefront of Carpl is Vidur, an entrepreneur with a unique blend of medical, technical, and business-building expertise. His decade-long experience in medical imaging, first with Mahajan Imaging and then with Carpl, Vidur possesses an unparalleled understanding of the medical imaging industry. His visionary approach has been instrumental in crafting Carpl’s value proposition, effectively addressing the intricacies of AI integration in the field of radiology.

The remarkable progress Carpl has achieved underscores its commitment to navigating the complexities of AI integration in radiology. Some of its customers include marquee names such as University Hospitals Cleveland, Albert Einstein Hospital Sao Paulo and Singapore General Hospital. By enhancing the role of front-line radiologists, Carpl is not only making significant strides in the industry but is also poised to bring AI adoption in medical imaging to the forefront. As partners in this promising journey, we are thrilled to collaborate with Vidur and his team, supporting them in their pursuit of revolutionising medical imaging through AI.

Read more on how Carpl is accelerating AI adoption in radiology on TechCrunch, Forbes, Bloomberg‍

Demand for software is growing twice as fast as the supply of software developers. As every business strives to become a software business, the demand will grow even more rapidly. However, there are only so many developers that the world can produce every year, leading to a continuous need for an increase in developer productivity. Front-end developers are no exception to this, and in many ways, the need there is even more pronounced than in other areas.

Front-end development automation isn’t a niche; it’s a sprawling, horizontal market. According to Stack Overflow, approximately 75% of all software developers work in front-end development, representing around 20M developers globally. Projections suggest this number will soar to approximately 33.5M by 2030.

Front-end developers play an important role in crafting user experience. With increased competition, rising user expectations and more complex applications, front-end development has become increasingly more complex in recent years. At the same time, developers continue to devote a significant portion of their work, ranging from 25% to 75%, to procedural coding tasks like styling (CSS), structuring (HTML), and the application of framework-specific boilerplate.

Now, imagine a world where front-end developers can focus exclusively on the core business logic of an application while automated tools handle the rest. This is the vision of Kombai that is seeking to revolutionise front-end development by reducing 60-70% of a developer’s workload. Their strategy involves automating the more repetitive aspects of front-end development, leaving developers with more time to concentrate on essential coding tasks.

Modern front-end coding process consists of two main parts – building the visual appearance that matches a given design and building functionalities (aka “writing UI code”) and interactions around the visual appearance-related code. To write UI code, developers rely on their experience, skills, and intuition to make numerous inferences, both explicit and implicit. Kombai’s proprietary ensemble model is able to “look at” complex, real-world designs, derive these inferences as a developer would, and generate high-quality code using that “understanding”. The model doesn’t require users to manually pre-process the designs. Further, Kombai is highly steerable – it can be nudged to generate the exact code a developer wants by “prompt engineering” the designs.

During our diligence, we discovered a large number of tools for automating front-end development. However, most of these fall woefully short of expectations. They are simply not production ready. Most tools today fail to interpret the designs intelligently and generate code that is not intuitive for developers to understand and maintain. It is a hard problem, and this is precisely what we like about it.

A crucial distinction lies in the company’s mission – they aim to “augment” developers, and make frontend development enjoyable for them, rather than replace them. Unlike no-code tools such as WordPress, Bubble, Retool, Builder, and Wix, which cater to non-developers creating lightweight and standardised apps, their innovation targets feature-rich and highly interactive applications. We believe that as the complexity and demand for front-end development increases, intelligent automation will become a necessity rather than a luxury.

In Dipanjan and Abhijit, we discovered a team with a unique combination of technical prowess and product acumen necessary to address the issue through a fundamentally fresh perspective. Right from the outset, it was evident to us that this team possessed a rare combination of individual excellence and functional harmony. In a relatively short period of time, they have been able to build a new model, not just a thin product layer on top of existing models, that can solve a globally-unsolved problem. We are delighted by the remarkable progress the team has made and the encouraging feedback they have got from developers during their private research preview phase.

We, at Stellaris, are excited to seize the opportunity to partner with the founders as well as with Foundation Capital, our co-investors in this company, on this promising journey.

Kombai is launching their public research preview today (23rd August, 2023). Do check them out at Kombai.com/launch.

More on our investment in Kombai in Techcrunch and Economic Times‍