Julius.ai

Julius.ai is an AI data-analysis company founded by Rahul Sonwalkar, its CEO. Its product analyzes and visualizes datasets and performs predictive modeling from natural-language prompts. Julius concentrated on people who work with data repeatedly, pairing generated analysis with controls for inspecting code and refining results.[2][4][1]

The company’s origins precede its data-analysis product. Julius participated in Y Combinator’s Summer 2022 batch, and YC lists its founding year as 2022. Sonwalkar launched Julius after leaving the logistics startup he had been building during the accelerator program. In a February 2024 anniversary demonstration published by Latent Space on March 9, he said the product had launched about six months earlier and reached 300,000 users, with several thousand using it daily.[2][4][1]

That demonstration used approximately 20,000 rows of country-level human-height data spanning 100 years. Sonwalkar explained an iterative execution loop: each user received a cloud virtual machine, where the AI wrote and ran code in a Jupyter notebook, inspected the output, and wrote further code as needed. Loading the demonstration file involved importing pandas and returning sample rows. He also described recovery when generated code failed on its first attempt.[1]

Sonwalkar attributed the choice of Python to its capacity for deeper analysis, including regression, and its ecosystem of numerical and visualization modules. He contrasted that flexibility with the difficulties his team encountered using SQL alone. The product thus used natural-language requests to drive executable analysis rather than requiring users to write every query themselves.[1]

Asked by Jerry Liu how Julius differed from ChatGPT’s Code Interpreter, Sonwalkar emphasized specialization in daily data work. He described collecting recurring errors from usage and changing rules and prompts to prevent them before execution. Users could reveal and edit generated code, give natural-language instructions about code, or request chart changes—for example, turning a line graph into a bar chart. These controls supported the final revisions needed to finish an analysis.[1]

Harvard Business School integrated Julius into its proprietary AI platform for the required MBA course Data Science and AI for Leaders. Iavor Bojinov asked Sonwalkar to modify Julius specifically for the course. Co-led by Bojinov and Karim Lakhani, the course was developed with Digital Transformation (DTx), the Digital Data Design Institute (D³), HBS IT, and the MBA Program. Students could request regressions and visualizations in ordinary language without writing code.[4][7]

In its March 28, 2025 account, HBS reported that Julius had engaged all 930 students and processed 126,704 messages by the end of the first course. Faculty could review anonymized interactions to track progress and adjust lessons. Julius operated alongside a separate DSAIL Tutorbot developed by DTx.[7]

On July 28, 2025, Julius announced that it had raised a $10 million seed round led by Bessemer Venture Partners. TechCrunch’s report named Horizon VC, 8VC, Y Combinator, and AI Grant as participants, alongside angel investors including Aravind Srinivas, Guillermo Rauch, and Jeff Lawson. The company said it had more than two million users and reported more than ten million visualizations; those figures did not specify paying customers or a measurement window.[4]

Jaden Moore’s June 5, 2025 Notebooks announcement advertised running selected parts of an analysis without executing the entire notebook again. Julius subsequently announced that it was sunsetting Notebooks and concentrating on its core chat experience for complex, multi-step analysis.[5][6]