01
Heads of data and analytics
Accountable for a lakehouse that is growing faster than the design that started it.
Fractional AI Architect / Databricks
Databricks rewards good architecture and punishes the other kind with a bill. The architect works next to your engineer on the decisions that set both: layout, governance, pipeline patterns and cost.
Databricks
Lakehouse and pipeline design
14 hrsPairing with your engineer
12 hrsGovernance and cost review
8 hrsDocumentation and handover
6 hrsThe decisions that matter on Databricks are made early and paid for monthly. How the medallion layers are actually drawn, what Unity Catalog governs and who grants it, whether pipelines are jobs or Delta Live Tables, how models get served and evaluated, and which compute choices are quietly costing you multiples of what they should. An architect part time is usually a better answer than an engineer full time, because these are judgement calls rather than volume work.
Who it is for
Teams with Databricks already in production, or about to be, and one engineer carrying more architectural weight than is reasonable.
Accountable for a lakehouse that is growing faster than the design that started it.
Building pipelines daily and making architecture decisions weekly, without a second opinion available.
Seeing a Databricks bill grow faster than the workload, and needing to know which choices caused it.
Coverage
The parts of a Databricks estate where an early decision compounds, for better or worse, every month afterwards.
Where bronze, silver and gold boundaries actually fall for your data, rather than the diagram version.
Governance that holds across workspaces: catalogues, lineage, grants, and who is allowed to issue them.
Partitioning, liquid clustering, file sizing and the maintenance jobs that keep read performance from decaying.
Jobs against Delta Live Tables, orchestration boundaries, idempotency and how failure is meant to behave.
Model registry discipline, serving topology, and the evaluation that has to run before anything is promoted.
Compute policies, warehouse sizing, job clusters against all-purpose, and where the spend is genuinely going.
Straight answer
An honest read matters more here than on most platforms, because Databricks is easy to adopt for workloads that never needed it.
Outcomes
Layer boundaries, governance and pipeline patterns written down and being followed.
Spend attributed to workloads, with the specific choices driving it identified and the worst of them fixed.
The same deliberate goal as every track: your team carrying more of it each month.
Working with Indian enterprises
Most of our engagements run with banks, NBFCs, insurers and manufacturers headquartered in India. The architect works IST, joins your existing rituals, and is used to the approval chain an Indian enterprise actually has.
Architecture reviews, vendor selection and security sign-off go faster face to face. The architect travels to your offices across the metros and tier 2 cities as the engagement needs it.
Data residency, consent and purpose limitation under the DPDP Act 2023 shape the architecture from the first session, alongside RBI, IRDAI and SEBI expectations where they apply.
Design decisions, model choices and control gaps are written down as you go, in a form audit and risk can read without a translation layer.
Where a workload cannot leave the country, the architect can design against open-weight models running entirely on Indian infrastructure through Gyde Inference.
Delivered onsite in
Free download
A sample engagement plan for a 40 hour month: what the architect does in week one, what your engineer owns by week four, and the artefacts that exist at the end of it.
Questions
If your question is here in a form we have not covered, ask us directly and we will answer it plainly.
They set the lakehouse and pipeline architecture, review what your engineer builds against it, and own the decisions that are expensive to reverse: layer boundaries, Unity Catalog governance, table design, serving topology and compute policy. Roughly a third of the time is spent pairing rather than reviewing.
Forty hours a month, about ten a week, split across design work, pairing with your engineer, governance and cost review, and documentation. The first month is heavier on architecture because that is when the target design gets set, and lighter on review.
Usually yes, and it is often the fastest visible return. Most overspend traces to a small number of choices: all-purpose clusters doing job work, warehouses sized for a peak that no longer happens, unmaintained tables, and pipelines rerunning more than they need to. Cost review is a standing part of the monthly rhythm.
Yes, and that is deliberate. The engagement is built around one dedicated internal engineer who does the building while the architect sets direction and reviews. Without that person the knowledge has nowhere to land and you are buying a dependency rather than a capability.
A consultancy typically staffs a team and delivers a project, then leaves. This is one senior architect on a standing monthly basis whose explicit goal is to make your engineer capable enough to need them less. It is a smaller commitment and a different measure of success.
All three. The lakehouse decisions are largely consistent across clouds, and the differences that matter are in identity, networking and how storage is governed. If you are on Azure, the same engagement can cover the Microsoft integration surface alongside it.
Yes, and it is where most of our work sits. The architect works IST and travels to your offices across Bengaluru, Mumbai, Delhi NCR, Pune, Hyderabad and Chennai. Data residency under the DPDP Act 2023 and RBI or IRDAI expectations are designed for from the first session.
That is the best time. The decisions that are expensive to unpick, layer boundaries and governance in particular, are all made in the first few months. An architect at the start costs less than a migration at the end.
Keep going
Start the conversation
Tell us the platform and the workload that is stuck, and we will propose a scope for the first 40 hour month.