The first useful description of Lightdash did not come from its founders. Hamzah Chaudhary and Oliver Laslett posted their young analytics tool to Hacker News with a title so careful and technical that Chaudhary later struggled to remember it. An early commenter translated the pitch into four economical words: “open source Looker.” Chaudhary’s reaction was immediate. That was a much better name for what they were doing. For roughly the next year, the startup ran with it.
It is a tidy anecdote about positioning, but it also captures Chaudhary’s style. He is an engineer turned product leader who listens for the plain sentence hiding inside a complicated system. His company sits in one of technology’s densest districts, business intelligence, where raw warehouse tables must become definitions, charts and decisions. His recurring question is practical: how can more people use data without making the numbers less trustworthy?
Lightdash’s answer has changed shape. It started as a tool for visualising a dbt project. It became a hosted BI service, then a semantic layer, an AI analyst and a place to build custom data apps. The underlying argument has stayed steady. Data teams should be able to work with the habits that make software teams effective: version control, testing, preview environments, code review and reusable logic.
“The mission from day 1 is to try and empower data people to be able to work more like software engineers.”Hamzah Chaudhary, Scaling DevToolsOrigin story / January 2020
The tool beside the business
The path began before Lightdash had its name. Chaudhary had studied European and Middle Eastern languages and linguistics at University College London, focusing on French and Arabic, before completing an immersive software engineering program at Hack Reactor. His early résumé wandered productively: Deloitte, a short spell at McKinsey, coding instruction, an NGO called WaterForAir and a late-night baked-goods delivery service in London. He describes himself in one old founder profile as an “average chef.”
At Cytora, the insurance technology company, the strands converged. Chaudhary joined as its second engineering hire and eventually became responsible for data pipelines and analytics. The software side of the business had mature ways to ship. The data side felt, in his telling, like a wild west of CSV files and documents saved to Google Drive. Concepts normal to developers, such as Git, development environments and production previews, were not normal for analysts.
He met Laslett at Cytora. When the pair left, they expected to bootstrap a data consultancy. The timing was January 2020. One assignment connected to the United Kingdom’s Department for Education involved aggregating viewing data for online curricula as schools moved onto the internet. They needed business intelligence for their own work, but did not want the bill from one of the large platforms. So they built a small internal tool.
A customer noticed it. “What’s that?” was the question Chaudhary remembers. The product looked rough, he conceded, but the customer wanted to use it. More people pulled at the same thread. The founders kept insisting the consultancy was the real business until the evidence made that position awkward. The consultancy stopped. Lightdash carried on.
Useful before it was comfortable
The first public release was barely collaborative. Chaudhary recalled that users could not even invite another person into the application. Still, some people tried to run it in production. That willingness mattered more than a launch-day spike. It suggested that the problem was painful enough for users to work around an unfinished product.
Open source then created a flood of possible directions. Chaudhary’s response was focus. The team had a specific user in mind: the data professional serving everyone else in an organisation. Because the target stayed narrow, requests tended to collect in recognizable buckets. Community feedback was evidence, while the founding idea remained the filter.
That idea also gave Lightdash a distribution model. A small team or hobbyist could self-host the complete open-source product. A larger company could run a proof of concept before security reviews and procurement, then choose the cloud product for managed deployment, performance and additional controls. Trying and buying became separate events. By October 2024, Chaudhary said more than 5,000 teams were self-hosting the software.
The commercial milestones followed the product proof. Lightdash announced $8.4 million in seed financing in 2022, composed of a $6 million Accel-led round and an earlier $2.4 million Moonfire-led round. In October 2024, it added an $11 million Series A led by Accel, with Shopify Ventures and Operator Partners joining and Y Combinator and Moonfire participating again. At the time, the company said two queries were being answered on the platform every second.
Trust before conversation
The arrival of generative AI gave Chaudhary’s old concern a new consequence. A person might tolerate a slow dashboard request. An agent can produce an incorrect answer instantly and at scale. Before it can answer “What was revenue last month?”, it needs a shared definition of revenue, knowledge of the right tables and permission to see them.
Lightdash’s semantic layer is meant to provide that grammar. Business concepts such as revenue, active users or churn live alongside a dbt project in readable, version-controlled files. The same definitions can power the interface, APIs, embedded analytics and AI. When Lightdash launched its AI Analyst with the Series A, Chaudhary emphasized that it used the existing query engine and permission system. Customers could choose a model provider, and much of the work relied on metadata rather than raw customer data.
“Don’t force people to use the tool who shouldn’t be.”Hamzah Chaudhary, on meeting users where they work
That line explains the broader move. Chaudhary has argued that people often want an answer inside Slack or another place where work is already happening. For a decade, BI products tried to draw users into their own interfaces. His wager is that governed data should travel. In August 2025, Lightdash open-sourced its semantic layer so those definitions could be used across tools, APIs and custom agents.
By July 2026, Chaudhary was writing about a step beyond the fixed dashboard. A marketing team may need a presentation, finance may need a forecasting model, and executives may need a report. Lightdash Data Apps let users describe a custom application in plain English while keeping its queries attached to the governed semantic layer. The visible thing can be bespoke. The definitions underneath stay shared.
Data-quality tests provide the starting point before the team changes direction.
The dbt-native BI product launches and finds an early community through Hacker News.
Natural-language analysis is attached to existing permissions and governed metrics.
Business definitions become reusable across interfaces, APIs and agents.
Custom data apps push the product beyond passive dashboard viewing.
The sentence beneath the system
There is an amusing continuity between Chaudhary’s linguistics degree and his work on semantic layers, though he has not publicly framed it as destiny. Both deal with meaning, context and translation. A warehouse column is not useful to most colleagues until somebody can say what it represents. An AI agent is not useful to a business until it can operate inside the organisation’s language.
His personal path also resists a polished founder myth. There were baked goods, an NGO, consulting, weekend coding classes, insurance models and a bootstrapped plan interrupted by a pandemic. Lightdash itself moved from consultancy tool to Hubble, from data-quality testing to BI, and from BI toward a more open substrate for applications and agents. Each turn kept a piece of the prior lesson.
Chaudhary’s ambition is therefore less about producing another screen full of charts. It is about increasing the distance that a data team’s judgment can travel. Define a metric once. Test it. Review it. Let a colleague explore it, an application use it and an agent discuss it without quietly changing its meaning.
The Hacker News commenter gave Lightdash its first concise category. Chaudhary’s work since then has been to make the category less confining. Open-source Looker was a useful way in. The larger idea is business intelligence that behaves like a well-made software system: inspectable, portable, governed and available wherever someone is ready to build.