Simon Eskildsen, a Danish-born software engineer based in Ottawa, is leading Turbopuffer Inc., a rapidly growing Canadian artificial intelligence startup that has developed a cost-efficient approach to vector search technology, a crucial component in AI data retrieval. Founded in 2023 alongside co-founder Justine Li, Turbopuffer addresses the escalating expenses associated with AI systems searching through large datasets, offering a solution that dramatically reduces operational costs for companies reliant on AI search capabilities.

The company’s origin traces back to Eskildsen’s move from Denmark to Ottawa in 2013 after being recruited by Shopify for an internship. Initially planning to spend only a gap year in Canada, Eskildsen decided to stay, drawn by the city’s environment and a sense of belonging within the local tech community. His collaboration with Li, whom he met at Shopify, brought together their complementary skill sets—Eskildsen’s deep infrastructure expertise and Li’s proficiency in programming—forming the basis for Turbopuffer’s technical development.

Turbopuffer’s technology centers on vector search, a method that converts text, images, audio, or any data type into high-dimensional numerical representations called vectors. These vectors are then indexed and searched to find items with semantic similarity, enabling AI systems to retrieve contextually relevant information from large volumes of data. Traditional vector search solutions primarily rely on expensive random-access memory (RAM) to process and store these vectors, contributing to high operational costs.

Eskildsen and his team innovated by designing an architecture that leverages object storage—an inexpensive form of data storage—as the primary medium for vector data. Their system dynamically loads relevant data into faster cache and RAM layers only when necessary, balancing cost savings with necessary performance. This architecture reportedly reduces vector database expenses by up to 70 times compared to conventional methods. While initial query speeds may be slower than fully RAM-based systems, subsequent queries benefit from caching, matching or exceeding competitors’ performance.

Starting with less than US$1 million in funding and a lean team of 37 employees, the majority of whom work remotely outside Canada, Turbopuffer has rapidly expanded its client base to approximately 1,200 customers, including prominent AI organizations such as Anthropic, the company behind the Claude AI system, and other tech enterprises like Cursor, Harvey, and Legora. The company is on track to exceed US$100 million in revenue this year and has already achieved profitability, distinguishing itself from many AI startups that require tens or hundreds of millions in venture capital.

The startup’s core customers are largely AI companies burdened by steep infrastructure costs but Turbopuffer’s technology is gaining traction beyond this sector. Enterprises with extensive datasets, such as telecommunications provider Telus Corp., have adopted Turbopuffer’s solution due to its affordability and scalability. Telus reportedly saw vector search costs drop from around $200,000 per month to $5,000 after using Turbopuffer’s system.

Despite its Canadian roots and Ottawa headquarters, Turbopuffer operates with a global workforce, reflecting the difficulties of attracting and retaining advanced AI talent locally. Eskildsen is actively encouraging employees to relocate to Canada and is advocating for policy measures such as improved immigration pathways, increased funding for math and science competitions, and enhanced connectivity with U.S. tech hubs to bolster Canada’s position in the AI sector. He underscores the importance of competitions like the International Olympiad in Informatics, which played a formative role in his own development and continues to be a source for hiring exceptional talent.

While Turbopuffer’s approach is recognized by experts as a pragmatic assembly of existing ideas rather than a wholly novel invention, its execution and cost-effectiveness have enabled it to carve out a distinct niche in the AI infrastructure market. Competition remains fierce, with major players like Amazon Web Services introducing similar vector search services. Still, Turbopuffer’s founders express confidence in their ability to maintain a competitive edge through continuous technical refinement and a focused product strategy.

Eskildsen, who has declined to relocate to Silicon Valley despite pressure, sees his company as a testament to Canadian innovation with a "strong Canadian heart." Though the company did not receive government funding despite applying for national AI initiatives, Turbopuffer’s success has been driven largely by entrepreneurial initiative and fortuitous circumstances. It stands as an example of Canada’s evolving role in the global AI landscape, balancing challenges related to talent attraction and competition with opportunities afforded by a vibrant tech ecosystem.