In the rapidly evolving world of enterprise AI, few challenges loom more prominent than the hunger for high-quality data. Former Google engineer and AI expert Pratik Karki is stepping up to meet this need with his latest venture, Anthromind, a stealth-mode startup making waves in the B2B SaaS space. Hailing from Kathmandu and an alumnus of Grinnell College and the University of Iowa in America, Karki brings both technical expertise and business acumen to this ambitious new project.
During his tenure at Google, Karki was pivotal on a pioneering Cloud and AI engineering team, helping deliver cutting-edge solutions to key large language model (LLM) accounts such as Anthropic, Character.ai, and Midjourney. His involvement spanned everything from Data Warehouse migrations to GPU/TPU strategy, placing him at the forefront of the AI boom and exposing him to the industry’s most pressing demands. “The greatest challenge we faced was access to top-quality data,” Karki recalls. “As models grew more sophisticated, the need for specialized datasets soared, yet the market lacked scalable, high-quality options.”
Sensing an opportunity, Karki co-founded Anthromind in 2024 with the goal of shaping the next generation of data infrastructure for enterprises building and fine-tuning LLMs. Although the company remains in stealth mode, its mission is clear: to offer best-in-class synthetic data and expert-vetted datasets that empower AI development across various industries, including healthcare, finance, and insurance.
Anthromind’s primary offerings include a synthetic data platform employing agent-based data flows and self-play techniques to produce realistic, domain-specific datasets. It also offers human-in-the-loop validation by in-house PhD experts, ensuring data accuracy and alignment. Citing Nat Friedman’s (former Github CEO) observation that “experts are the new GPUs,” Karki’s team integrates expert oversight into its data processes, recognizing that specialized human knowledge remains critical to top-notch data quality.
Already, Anthromind has begun working with multiple enterprise customers, including foundational AI research organizations, and has garnered significant interest from cloud security and fintech players. Having closed a funding round at the end of 2024, Karki hints at further developments in early 2025. “We’re excited to announce more progress soon,” he says, noting that Anthromind occupies a strategic point in a field once dominated by established data labeling giants like Scale AI.
Beyond just addressing a gap in AI data, Karki believes Anthromind could define the future of scalable oversight for powerful AI models. “As AI models become more powerful, the notion of weaker AI systems aligning super-powerful models is fascinating,” he explains. “Anthromind is positioning itself to lead that charge, focusing on both automation and human expertise.”
For Karki, transitioning from a tech giant like Google to launching Anthromind was a logical next step. “The opportunity ahead is enormous,” he says. “We see ourselves not simply as a vendor of datasets but as a driver of how enterprises build, fine-tune, and trust their LLMs. The coming years in AI will be thrilling, and we plan to be right at the frontier of this new data infrastructure era.” #Nepal
