Numbers Station Empowers Business Users to Easily Communicate with Their Data

Numbers Station, a cutting-edge startup leveraging large language models (LLMs) to enhance its data analytics platform, is excited to announce the launch of its first cloud-based product: Numbers Station Cloud, now available for early access. This innovative service empowers enterprise users to effortlessly analyze internal data through a user-friendly chat interface.

While many tools aim to transform natural language queries into database languages like SQL, Numbers Station's team contends that these solutions have significant limitations. The challenge lies in the generic nature of LLMs, which often lack insight into a company's unique operations, data structures, and internal terminology.

Co-founder and CEO Chris Aberger expressed his frustration with the common narrative around users being able to “chat with their data.” He emphasized that the real value lies in enabling non-technical users and business executives to ask meaningful questions and obtain answers from structured data sources. “Substantial data modeling and integration efforts are required around these foundational models and LLMs to ensure effectiveness,” he noted.

To achieve this, Numbers Station invested heavily in engineering a sophisticated component known as its semantic catalog. This catalog serves as an automatically curated repository of a company's specific metrics and definitions, tailored individually for each organization. Aberger described it as a “beastly thing” that aligns the model's understanding of terms like “recurring revenue” with each company's specific context.

While Numbers Station’s platform is built on advanced, specialized LLMs and machine learning frameworks, it is the semantic catalog that acts as the backbone of the system. Co-founder and chief scientist Ines Chami acknowledged that the team initially underestimated the complexity involved in developing this aspect of the platform. “It involves classical machine learning and data engineering principles: How can we create a representation of knowledge that the model can utilize effectively?” she explained. The goal is to transform vague inquiries into clear, actionable queries. Numbers Station’s findings indicate that this methodology significantly enhances accuracy compared to traditional text-to-SQL pipelines.

Despite the rollout of this chat feature, the company’s vision extends far beyond its current offering. Aberger stated, “What we’re fundamentally building is an AI platform for analytics. This chat service is just one application of the broader objectives we’re pursuing, including addressing various data challenges such as enriching data with third-party sources and applying established algorithms like fuzzy matching. The possibilities for expansion are nearly limitless.”

Numbers Station has already secured partnerships with several Fortune 500 companies, including the prominent global real estate services firm, Jones Lang LaSalle. “Numbers Station stands at the forefront of enterprise AI for structured data,” remarked Sharad Rastogi, CEO of Work Dynamics Technology at Jones Lang LaSalle. “We are impressed by Numbers Station’s reliable and engaging platform, which learns continuously as we utilize it, allowing our data teams to explore and validate hypotheses that drive impactful business outcomes.”