# Databricks implementation, from lakehouse to live application > If you already run Databricks, there's usually one initiative that keeps slipping, often a migration off Snowflake or Synapse that stalled halfway while your own engineers were needed elsewhere. We pick it up and finish it in your workspace, and the people who scope the work are the people who build it. Eighty of our engineers hold Databricks certifications, inside a team of 115+ across four offices, in Gilbert, Amsterdam, Dnipro and Hyderabad. Once the data has moved, we design Unity Catalog so grants actually hold and ship the applications and agents that run in the workspace under their own service principal. Canonical: https://www.techfabric.com/expertise/databricks --- Eighty Databricks-certified engineers. Applications that run in the workspace, and the pipelines that feed them. ## What we build - Databricks Apps and Lakebase services, built for in-workspace deployment - Governed agents that inherit Unity Catalog permissions instead of routing around them - Model serving, Unity Gateway, and evaluation gates on promotion - Warehouse and ETL migration with reversible cutover, when that is the way in ## Migration Off Snowflake, Synapse, Teradata and SQL Server. The conversion is the part everybody plans for and the part that goes fine; what sinks a migration is being unable to prove afterwards what was in scope. - Twenty-seven source platforms with executable playbooks rather than a methodology deck - Signed certificates per object recording what converted, by which tool version, and what validation said - Waves rather than a big bang, each validated and each reversible - Decommissioning as a dated phase, because otherwise both bills run for a year ## Unity Catalog and governance Most rollouts stop halfway. The catalogues exist, half the estate is still external tables against the old metastore, and there is a service principal everybody has stopped asking about. - Catalogue and schema design shaped around the organisation rather than the storage layout - Grants expressed through groups, so joining a team is what changes access - Hive metastore migration in stages, with external tables where a full move costs more than it returns - Lineage that survives the transformations you actually run ## Lakeflow and pipelines The pipelines and the operational databases underneath them, built so the numbers reconcile and a replay rebuilds a view rather than approximating it. - Medallion structure designed against how the data is queried, not how it arrives - Streaming and change data capture where the business needs the number now - Lakebase for the transactional side, with the sync direction and freshness stated rather than assumed - Warehouse sizing, scheduling and layout, which are the three levers that decide the bill ## Databricks Apps and agents The part most partners stop before. A platform gets used when somebody builds the application on top of it, and that is application engineering rather than data engineering. - Apps running in-workspace under their own service principal - Agents on Unity Gateway inheriting catalogue permissions instead of routing around them - Model Serving and the evaluation that says whether a change helped - Genie semantic layers that hold up when finance checks the answer ## The bench Eighty Databricks-certified engineers inside a team of 115+. That is the number worth asking every partner for, because it is checkable and a partner tier is not. - Certifications held by individuals, not a company badge - Average fifteen years of production experience per engineer - The people who scope the work are the people who build it - Four accelerators of our own, open source, that you can read before hiring us ## Questions ### We already have a Databricks team. Where do you fit? Usually on the initiative that keeps slipping because your team is fully committed elsewhere: the app that never leaves the notebook, the agent nobody trusts with production data, the Lakebase service that keeps getting deferred. We take ownership of that piece without pulling anyone off the current roadmap, and we work in your workspace so nothing has to be handed back later. The Databricks Health Check at /databricks/health-check is the two-week way in when the question is the bill rather than a new build. ### Do you build Databricks Apps, or just pipelines? Both, and the applications are the part most partners stop short of. Applications that run in-workspace under their own service principal are a large part of what we do. TechFabric Runway at /accelerators/runway is the delivery accelerator we use when a team has to ship those apps on a cadence, with preview environments and evaluation gates on promotion. ### How do you know an agent or a Genie answer is any good? An evaluation harness with ground truth you own. TechFabric Experiments at /accelerators/experiments keeps that suite running against your own Databricks data, so a failed evaluation blocks promotion. When executives have already stopped trusting Genie, the named engagement is Genie Accuracy at /databricks/genie-accuracy. ### How many Databricks-certified engineers does TechFabric have? Eighty, as things stand today, sitting inside a team of 115+ across four offices, in Gilbert, Amsterdam, Dnipro and Hyderabad. Certification is the floor. Fifteen years of average production experience is what decides whether the thing still works in month six. The named ways to start are on /databricks. ### Can you migrate our warehouse to Databricks? Yes. The first two weeks are the Migration Readiness Sprint at /databricks/migration-readiness: inventory, exclusions, a wave plan, and a go or no-go. Underneath it is TechFabric Airlift at /accelerators/airlift, which composes Databricks Lakebridge and wraps conversion in signed certificates and a reversible cutover. Twenty-seven source platforms have executable playbooks. ### Do we have to already be on Databricks to work with you? No. Some engagements start before the first workspace exists, some rescue a stalled pilot, and some are a migration off a warehouse that has stopped paying for itself. The six-week Lakehouse Launchpad at /databricks/launchpad is the named way to stand a governed lakehouse up. We also build on Azure, AWS, GCP and Cloudflare.