AI has showcased remarkable capabilities, but the challenge remains in determining which tasks are best suited for these technologies. Reliant, a new startup, is focusing on alleviating the tedious and time-consuming tasks that often plague research and academia, particularly the extensive data extraction work traditionally handled by graduate students and interns.
Karl Moritz Hermann, CEO of Reliant, highlights the company’s mission: “The best thing you can do with AI is improve the human experience: reduce menial labor and let people do the things that are important to them.” In the research world, literature review is one such menial task that consumes significant time and effort. Researchers often sift through thousands of papers to extract relevant data, a process ripe for automation.
Hermann recalls a study where the authors reviewed 3,500 scientific publications, with many turning out to be irrelevant. This labor-intensive process, extracting only minimal useful information, seemed an ideal candidate for AI automation. While language models like ChatGPT have attempted this task, they often fall short, with an error rate that’s far from acceptable in critical research environments.
Reliant’s core product, Tabular, is built on a combination of a language model (Llama 3.1) and proprietary techniques, making it significantly more effective than current tools. For instance, in the aforementioned study involving thousands of papers, Reliant’s system completed the task with zero errors. The tool allows users to upload documents, specify the desired data, and Reliant extracts the information accurately, regardless of how well it’s labeled or structured.
“Our users need to be able to work with all the data all at once,” says Hermann, emphasizing the importance of allowing researchers to focus their attention where it’s most needed. Reliant’s system not only extracts data but also enables users to edit it and navigate back to the source literature as needed.
This meticulous approach to AI application in research has caught the attention of investors, leading to a successful $11.3 million seed round led by Tola Capital and Inovia Capital, with participation from angel investor Mike Volpi.
Reliant’s strategy includes owning its hardware to handle the intensive compute requirements of its AI models. This in-house approach allows the company to pre-process data, predicting and preparing answers to common research queries, which significantly enhances efficiency.
The startup’s focus on precision and reliability positions it well in a market where mistakes are costly. Unlike broader AI applications that may tolerate occasional errors, Reliant aims to serve domains where accuracy is paramount. As Hermann notes, “We’re for where precision and recall really matter, and where mistakes really matter.”
As the biotech and research industries increasingly integrate AI, Reliant is poised to play a crucial role, providing tailored, reliable solutions that enable researchers to focus on innovation rather than drudgery.
