"Data readiness" is a Myth: Reliable AI with an Agentic Semantic Layer — Anushrut Gupta, PromptQL
AI Engineer World's Fair 2025 · 17:02
Enterprise AI workspace and data analytics
PromptQL builds an AI workspace for teams to analyze business data, generate dashboards and reports, and automate recurring work. It connects databases, warehouses, SaaS applications, and APIs without moving or reshaping their data, then writes code to answer questions across those systems. Business and technical teams can review answers in shared threads and maintain a collaborative wiki of organizational context. Developers can also embed PromptQL in customer-facing products or call it from other agents. Generated programs run in a sandbox, with source permissions including row- and column-level controls.
Founded as Hasura in 2017 by Tanmai Gopal and Rajoshi Ghosh, the company rebranded around PromptQL in 2025, with Gopal as CEO. Its engineering roots include the Hasura GraphQL Engine and Data Delivery Network. PromptQL separates AI planning from programmatic execution; its research explores a domain-specific language for verifiable execution plans and a Domain Learning Layer that captures organizational knowledge from live systems, repositories, and user interactions.
By 2022, the company reported more than 400 million Hasura downloads. Hasura raised a $100 million Series C led by Greenoaks in 2022, bringing cumulative funding to $136.5 million at a reported $1 billion financing valuation.
AI Engineer World's Fair 2025 · 17:02
Affiliations reflect their AIE appearances, not necessarily current employment.
Gupta presented an agentic semantic layer in which an LLM generates PromptQL plans for a deterministic runtime to execute across distributed data sources. Examples showed table discovery and adaptation to company-specific terminology and analytical requirements.
Affiliations reflect each recorded session, not necessarily current employment.