The glamorous part of enterprise AI is the answer appearing on a screen. The less photogenic part is deciding what the answer means. Somewhere behind the demo, a company must agree on what counts as revenue, which customer record is current, and whether yesterday’s numbers actually arrived. Infinite Lambda has built its business around that work. Its bet is that the road to useful AI runs through the data systems companies have spent years postponing fixing.
There is a striking example at AstraZeneca. In a testimonial published by Infinite Lambda, enterprise data director Greg Holland says modernisation helped reduce clinical trial data submission time from 30 days to two. He also reports millions in licence savings. Those are customer-reported outcomes, rather than a forecast for every buyer. But they give the company’s pitch something more substantial than a chatbot screenshot: an existing process, an elapsed time, and money released from old software.
A consultancy with a software engine
Founded by Nas Radev and Márton Hubay, Infinite Lambda was incorporated in Britain in April 2019. Radev is CEO; Hubay is co-founder and CTO. The London-based business sits in the data and AI services market, where organisations hire specialists to plan architecture, build platforms, move workloads, and teach employees how to use the result. Its website describes a team of more than 100 people, spanning engineering, consulting, analytics, and delivery.
Its current centre of gravity is Flowline, a proprietary modernisation offering. The company calls its approach “Services-as-Software”: packaging engineering methods and automation into delivery work that would otherwise consume substantial consultant time. Buyers are purchasing a transformation of their own systems, with implementation and training attached. The commercial proposition includes fixed-cost, fixed-timeline engagements, a detail that matters to executives who have watched migration schedules become a permanent item on the agenda.
Holland’s account helps explain what can change a buyer’s mind. He points to previous problems the team had solved, confidence in the schedule, and a fixed-cost contract. The persuasion rested on execution being legible. A transformation becomes easier to sponsor when the destination, duration, and commercial exposure can be discussed together.
“saving us millions in licence fees.”Greg Holland, Enterprise Data Director, AstraZeneca, in Infinite Lambda’s customer testimonial
The old system sent the first invoice
In one insurance case study, Infinite Lambda describes a business serving more than six million members whose acquisitions had left a patchwork of on-premises systems. Platform costs were doubling year over year, while data requests could sit in a months-long queue. The pressure arrived through everyday operations before it became an AI question.
The response combined six weeks of discovery and DataOps design with a move to BigQuery and dbt Cloud. Infinite Lambda reports 95% automated pipeline conversion, completion of a two-year roadmap in under six months, and €800,000 in annual licence savings. Analysts also received dbt training. The useful financial distinction is between recurring savings and project cost: retired licences are one component of the economics, alongside migration work and the replacement platform’s running costs.
Zego offers another version of the problem. Its Redshift warehouse had become expensive and difficult to scale reliably, according to Infinite Lambda’s account. The consultancy moved hundreds of sources and tables to Snowflake while preserving business continuity. Business and technical teams gained more direct access to data, reducing dependence on a single point of expertise. A cloud warehouse had itself become the legacy system. Being in the cloud, this case suggests, does not settle the architecture question forever.
Copy the checks, not just the code
Flowline’s migration method is organised around four stages: conversion, reconciliation, refactoring, and adoption. It maps dependencies in the old estate, translates workloads to tools such as dbt and Fivetran, checks converted data against the original at row level, and prepares the result for teams to maintain. Its supported legacy territory includes Informatica, Talend, SSIS, and stored procedures.
There is a practical lesson here for anyone commissioning a migration: make validation a deliverable. Code that runs is only part of the acceptance test. A business also needs confidence that totals, relationships, and downstream calculations survived the move. Dependency mapping matters for the same reason. A pipeline can be perfectly translated and still arrive in the wrong order for the systems that depend on it.
The company’s public GitHub gives engineers a smaller-scale way to inspect its interests. Repositories include dq-tools for working with test results, dbt-data-diff for comparing data in Snowflake, and dbt-tags for managing masking policies. These are useful clues to the craft behind the consulting pitch: checking, comparing, and controlling access. Readers can inspect that code before making any decision about a commercial engagement.
Your dashboard and your agent need a dictionary
Migration is also an opportunity to uncover what the business has been doing. Infinite Lambda’s ontology offering reads rules embedded in old pipelines and stored procedures, then produces a first-pass map of concepts and relationships. Think customers, orders, approvals, and the calculations that turn transactions into reported revenue. The organisation’s operating memory is often sitting inside code rather than a tidy handbook.
The qualifier “first-pass” is doing useful work. Extracting yesterday’s logic does not make it tomorrow’s policy. Infinite Lambda’s process includes refining that model with the client’s teams. In practical terms, software can expose two competing definitions of a customer; accountable people still have to decide which definition belongs in a particular decision.
The next service tackles business intelligence. Flowline inventories reports and dashboards, considers usage, and helps teams decide what to retain or retire. The destination is a shared semantic foundation using dbt metrics and Omni. That means the rules behind a number can be reused by a dashboard and an AI agent, instead of being quietly rewritten in each place.
There is an appealingly unromantic feature of this approach: some reports get thrown away. Moving every dashboard would preserve clutter as faithfully as useful knowledge. For a buyer, a sensible brief therefore includes the decisions each report supports and who relies on it. A migration can make an organisation’s reporting habits visible, including the expensive ones nobody would deliberately choose again.
At Skyscanner, the gap was between teams
Infinite Lambda’s Skyscanner case study starts with business domains developed rapidly and separately. Reports could not readily cross team boundaries, and cross-domain data products were blocked. The response began with an inventory of assets, processes, and internal champions, followed by a common strategy.
The resulting platform used dbt, Apache Airflow, Kafka, and AWS. Shared tooling was accompanied by data contracts and employee training. The lesson is organisational as much as technical: interoperability needs agreements about what teams supply and what other teams can expect. Buying common software does not, by itself, produce that agreement.
This helps locate Infinite Lambda among the alternatives. An internal engineering team can own the same work; a large systems integrator or another specialist data consultancy can deliver it. Infinite Lambda’s proposed distinction is the combination of repeatable automation, platform expertise, and adoption work. Snowflake and Fivetran are part of the delivery ecosystem, rather than substitutes for the whole service. Buyers should judge the fit against their actual estate and their team’s capacity to run what comes next.
Even a wildlife park has a data boundary
Mandai Wildlife Reserve, whose parks include Singapore Zoo, supplies a particularly concrete test of that fit. Its modernisation involved moving from SQL Server and SSAS Cube to dbt and Snowflake under strict security requirements. Infinite Lambda used Fivetran’s hybrid deployment, placing ingestion agents inside Mandai’s AWS environment. The architecture had to accommodate controlled access and automated, reviewed changes.
In its August 2026 award announcement, Infinite Lambda describes more than 40 pipelines ingesting over 270 tables daily for Mandai. It also describes a Claude Code agent limited to the infrastructure repository, with engineers reviewing and approving proposed changes. That is a useful picture of bounded AI deployment: a defined workspace and a human decision before changes take effect.
The implication for prospective clients is straightforward. A migration method must fit the organisation’s network rules, supported source systems, access permissions, and deployment controls. If those conditions prevent discovery or validation, the timetable has to reflect that. Likewise, automation cannot settle business definitions when the people responsible for them are unavailable. The prerequisites belong in the project design.
The people behind the pipelines
Infinite Lambda advertises local working hours, location flexibility, dedicated learning time, and unlimited paid holiday. Those are employer-stated policies. Its account of distributed working adds an operational detail: engineering chapters and project teams are organised across geography. A shared specialism can connect colleagues who do not share an office.
Expansion has followed that distributed model. In March 2025, the company announced its first teammates in Chile, Argentina, and Brazil. In May, it introduced Jim Ternyila to lead its US expansion. His connection with Radev predates the company: the two worked together at AstraZeneca in 2015. That earlier working relationship is a more revealing growth detail than another pin dropped on a map.
There are newer institutional markers, too. B Lab records Infinite Lambda as certified since April 2026, with an impact score of 91.2. The company announced ISO 27001 certification in July and Snowflake’s ASEAN Partner of the Year award in August. These credentials cover different ground: social and environmental assessment, information security management, and a technology partner’s recognition. They help buyers examine the business beyond a sales demonstration.
Its book, Data & AI: Fast and Slow, borrows inspiration from Daniel Kahneman to connect immediate operational decisions with slower analytical ones. That pairing also captures the company’s practical appeal. Organisations want faster answers, but they need the patient work that makes those answers dependable. Infinite Lambda sells the engineering, software, and instruction required to bring the two closer together.