Following Recent Biltmore Launch, Magic Data Automates the Messiest Parts of Data Analytics
Transforming raw information into actionable insights often requires weeks of tedious work or incurring hefty consulting fees; however, A new Atlanta startup is betting an AI “analyst” can handle the grunt work instead.
Mid-sized enterprises routinely spend hundreds of thousands of dollars on outside consultants to organize their data. Procurement and finance teams, in particular, account for 30% of annual analytics spent in the US, primarily on manual analytics processes. Pavleen Thukral believes there’s a better way. Based in Atlanta’s Tech Square, he founded Magic Data in 2024 to automate the messy data processing that bogs down spend analytics. “Data engineering always felt like the bottleneck of the companies I was building,” Thukral said, explaining why he set out to create a fix. His solution: an AI-driven tool that acts like a virtual data analyst, cleaning, organizing, and analyzing raw data so humans don’t have to.
An Application Layer on Top of AI Models
Magic Data is built as “the application layer on top of a foundational model,” Thukral explained. In practice, that means the startup isn’t inventing a new AI from scratch, but instead, harnessing the power of existing large AI models and adding a specialized interface for business data. The user experience is intentionally simple. “We want you, the executive, to chat with the AI and ask questions like normal,” Thukral said. Ask a plain-English question about your vendor data or customer records, and Magic Data’s software does the rest. “It’s up to our product to do the data discovery and analysis,” he said, describing how the tool automatically fetches and cleans the necessary data to inform financial officers and procurement managers on where they can save millions.
Behind the scenes, Magic Data connects to a company’s various data sources, from spreadsheets to data warehouses, with minimal setup. The core technology is uniquely focused on identifying and resolving data quality and integrity issues in real-time. Thukral says getting started is as easy as “[pointing] at a data resource,” with deployment taking as little as 10 minutes. Once connected, the AI begins scanning and understanding the data at a semantic level. It doesn’t just look for exact keywords; it recognizes concepts across different tables and databases, identifying links that might not be immediately apparent. This enables it to handle a level of complex, cross-dataset analysis that generic chatbots cannot perform natively.
Replacing the Data Janitor Work
Magic Data aims to replace much of the tedious, manual labor that data analysts and consultants traditionally perform. The platform automates data preparation, cleaning, and discovery — tasks that often consume most of an analyst’s time. “Magic Data does the work of a data analyst,” Thukral said plainly. It can, for example, find and fix messy entries that a human might spend hours hunting down. One illustration Thukral gives is if phone numbers are buried inside an unstructured text field of a database, Magic Data will automatically locate and extract them into the proper column, saving countless hours of cleanup. The software is also smart enough to ask the right questions about the data — much like a seasoned analyst would flag anomalies or identify interesting trends to explore. Addressing these types of quality issues can consume the majority of a team’s time or create a blocker altogether. Now, human analysts can spend their time on complicated areas that require business domain expertise. The rest of the questions can be handled self-service by the stakeholders who need them.
Thukral notes that data discovery, determining what data is essential and how to utilize it, is typically something companies pay consultants a significant amount of money to do. Magic Data’s system can perform data discovery on raw enterprise data, much like a top consulting firm, but at a fraction of the time and cost.
The tool is designed for mid-tier enterprises and above, targeting procurement and finance organizations. These are organizations that often have vast amounts of data and spend. The signal-to-noise is low. By automating much of the grunt work, Magic Data aims to relieve that strain and save these organizations millions of dollars. Thukral envisions the platform eventually functioning not just as an on-demand analyst but as a proactive aide. “I’m focused on making our tech proactive,” he said, imagining that it can “live almost as a teammate” alongside employees, monitoring data and surfacing ways for teams to save money or identify issues before anyone even thinks to ask.
From Hackathons to Tech Square to Fintech Exit
While Magic Data tackles the data challenges facing mid-sized enterprises, its founder’s path through Atlanta’s innovation ecosystem helped shape both the product and the company. Thukral’s journey runs through Tech Square and the broader Atlanta tech community.
As an undergraduate at Georgia Tech, he co-founded HackGT, one of the nation’s largest collegiate hackathons. Organizing such a massive event honed his leadership skills and plugged him into a network of ambitious technologists. “While at Tech Square, I met tons of CEOs and CTOs,” he said, describing how the district introduced him to many Atlanta tech leaders — connections that would prove invaluable when he later became a founder.
After earning his computer science degree and conducting research with TSRB’s Professor Thad Starner and the NSA, Thukral dove into building companies, including a fintech startup that created an online marketplace for banks to trade loans. The venture gained traction with major investors and was eventually acquired by a large financial technology firm.
That exit gave Thukral a taste of startup life and firsthand experience with the data bottlenecks that come with scaling a business. “From startups to enterprise-level companies, [I saw] time and time again that data engineering is often the bottleneck,” he said. With fresh experience and a Rolodex of talent, he set out to solve that problem through Magic Data. As a second-time founder, he assembled a seasoned team composed mainly of colleagues from past ventures. He secured early backing from local entrepreneurial legend Christopher Klaus to launch the startup in early 2024.
Magic Data spends a lot of its time in Tech Square, particularly around the Biltmore Innovation Center. “I love it here because we’re able to tap into a huge talent pool,” Thukral said. The density of ideas and expertise has already paid dividends, from connecting with skilled engineers to landing pilot customers. Thukral also credits the Tech Square community with helping him raise money and find mentors in the company’s early days.
Now that Magic Data is gaining momentum, Thukral is blending his experience as a hackathon organizer, researcher, and fintech executive to tackle a universal business challenge. The startup’s immediate focus is rolling out its platform to mid-sized and large enterprises eager to leverage their data more effectively.
Looking ahead, Thukral envisions Magic Data as an ever-present helper in a company’s analytics stack, quietly performing the heavy lifting and alerting teams to key findings in real time. In his words, Magic Data aims to function “almost as a teammate,” allowing business leaders to have a conversation with their data and get answers with no tedious preparation required.