There is a particular kind of software problem that looks small until you try to remove it. A developer needs data from a database. The request sounds like a sentence. The work becomes a week of endpoints, permissions, pagination, caching, deployment and the quiet terror of discovering that the supposedly temporary API has become important. Tanmai Gopal has spent much of his career staring at that gap between the simple request and the machinery required to make it safe.
His first durable answer was Hasura, the company he co-founded with Rajoshi Ghosh. Its open-source GraphQL Engine could sit over data and generate a usable API, turning a familiar pile of backend work into configuration and metadata. His current answer is PromptQL, an attempt to let AI work with company data and shared knowledge without treating security as an afterthought. The nouns changed. The irritation did not.
This makes Gopal's career easier to understand as a single argument than as a series of pivots. Software should give people more agency. Infrastructure should absorb repetitive complexity. Useful systems should respect the constraints of the organizations in which they live. He has been trying to reduce the distance between intention and execution, first for application developers and now for teams working alongside AI.
The market as laboratory
Gopal studied computer science at the Indian Institute of Technology Madras from 2007 to 2012, completing an integrated bachelor's and master's degree. His academic work included computer vision and machine learning, along with a research internship at Microsoft Research. A research career was plausible. He was drawn instead to the harsher feedback loop of making something and asking people to use it.
He skipped campus placements. The decision was not an abandonment of theory so much as a demand that theory survive contact with the world. Consulting followed. So did experiments in consumer software, including food delivery and food discovery. With 34 Cross, the software development and consulting company he co-founded before Hasura, Gopal worked on web and mobile products and helped large companies move from monolithic systems toward cloud-native architectures.
“Until you build a product and take it to market, you actually don't know anything.”Tanmai Gopal, discussing his pre-Hasura years
The sentence has the appealing severity of something learned through invoices. Consulting showed him where enterprise software became tangled. Consumer projects showed how engineering elegance could lose an argument to distribution or unit economics. Each encounter made the recurring data-access work harder to ignore. Teams were repeatedly building the same bridge between applications and databases, then maintaining it while both shores moved.
One problem, three forms
A quarter-million students and one stubborn idea
Before Hasura became widely associated with GraphQL, Gopal helped teach Introduction to Modern Application Development, or IMAD. The online course, created with IIT Madras, reached more than 250,000 learners. Conference biographies later described him, with disarming accuracy, as a “Stack Overflow powered fullstack, polyglot developer.” The phrase carries a worldview: expertise can be serious without being ceremonial.
Teaching at that scale reveals every unnecessary step in a tool chain. Beginners find the seams experts have learned to ignore. They meet setup errors before ideas, and boilerplate before feedback. Gopal's interest in education and his product instincts met at the same point. The best infrastructure does not merely run reliably. It lets more people participate without pretending that the difficult guarantees have disappeared.
Hasura turned that philosophy into a developer tool. Point the engine at a supported database and it generated a GraphQL API with relationships, filters, subscriptions and a permission model. The underlying engineering was not slight. Gopal wrote publicly about compiling GraphQL into efficient SQL, avoiding the familiar performance traps of naïve resolvers and keeping authorization in the query plan. The interface felt quick because the complexity had been moved, not wished away.
When the open-source engine arrived in 2018, adoption supplied unusually direct feedback. Developers could run it, inspect it and report exactly where reality disagreed with the roadmap. Hasura grew a commercial business around managed cloud and enterprise capabilities while keeping the project available to individual builders.
Community is a product constraint
Gopal has rejected the notion of open source as a clever acquisition trick. “It's not a hack,” he said while describing Hasura's community practice. Engineers and product leaders joined monthly calls. Users saw specifications, tried features and complained where the work needed improvement. That routine mattered because open source creates a constituency, not merely a download count.
The approach helped Hasura travel from side projects to mission-critical enterprise systems. It also created a useful tension. Individual developers wanted speed and a generous local experience. Larger organizations wanted network controls, governance, support and predictable operations. The commercial opportunity emerged from serving both groups without making either feel like bait.
In February 2022, Hasura announced a $100 million Series C led by Greenoaks. The round brought reported total funding to $136.5 million and a $1 billion valuation. Gopal's announcement spent less time admiring the number than mapping the work: expand the community, support more data sources, keep releases moving and strengthen the teams serving enterprise customers. Capital, in his telling, was permission to keep building the system, not an ending.
GraphQL meets the age of prompts
By 2025, the interface between people and computers was changing again. Hasura introduced PromptQL, which Gopal called the “spiritual successor to GraphQL for the age of AI.” Hasura GraphQL Engine and the newer Data Delivery Network would continue. PromptQL took the underlying data-access thesis into natural language, aiming to generate programs that combine retrieval with business logic while retaining controls around performance, deployment and security.
The ambition sharpened in 2026. Gopal began writing about “multiplayer AI”: shared agents that participate in a team's work rather than living inside one person's private chat window. He argued that ordinary workplace chat captures coordination more often than execution. The valuable context also sits in code, documents, data systems, decisions and the habits by which a team gets something done.
“The alpha has shifted from using AI to maintaining useful context.”Tanmai Gopal, 2026
There is a trap hidden inside that pleasant phrase. Shared context is useful only when it remains correctly shared. A finance leader, an engineer and a contractor should not receive identical answers simply because the agent can reach the same underlying warehouse. Gopal's writing on agent identity returns to authorization: whose credentials does the AI use, which context can it retrieve, and what may it reveal in a room containing several people?
PromptQL's 2026 product work placed those questions inside team collaboration. The company presented a shared AI coworker for its own workspace and announced PromptQL Tag for Microsoft Teams. The agent could be given an organization's name, models and tools while adapting its access to the identity of the person asking. It is an infrastructure founder's version of an AI product: begin with the permissions diagram, then allow the conversational magic.
The useful kind of impatience
Gopal's public voice mixes technical precision with impatience toward received wisdom. He can write architecture notes about query planning, then make a deliberately noisy product claim on social media. The playfulness is not separate from the engineering. Both are attempts to make an abstract problem vivid enough that people will argue with it.
His working relationship with Ghosh has also carried the company across several identities: consultancy, developer platform, GraphQL company, data-delivery business and now AI product builder. Their 2022 celebration photograph is gloriously unserious. Two founders grin at the camera with their fingers raised like horns. Startup mythology prefers the lonely visionary at a whiteboard. Actual companies are usually built by people willing to look ridiculous together after years of difficult meetings.
The ambition underneath Gopal's work has remained unusually legible. He wants technology to increase individual agency. In the Hasura years, that meant letting developers create and change applications without waiting on repetitive backend work. In the PromptQL years, it means letting people ask their organization's data to help make a decision or carry out a task.
The risk is equally legible. Every abstraction eventually meets the detail it hoped to hide. GraphQL APIs meet query cost and authorization. AI agents meet ambiguous language, stale context and confidential information. Gopal's bet is that these failures are invitations to build better infrastructure, not reasons to retreat into hand-written plumbing.
A career can be organized around a technology, and technologies age quickly. Gopal organized his around a complaint. Building useful software still asks too much repetitive work from people with better things to try. GraphQL supplied one answer. Shared-context AI may supply another. The durable asset is the willingness to keep asking the annoying question after the fashionable answer has changed.