Before I add to this, some context for why I'm weighing in: I've worked with Databricks for over 10 years and have deep respect for what the platform delivers โ Spark, Delta Lake, and MLflow are big open source contributions.
That respect is exactly why the training experience stands out to me, and why I want to see it raised to the same standard. So the notes below are meant constructively, from someone invested in Databricks succeeding โ not as a complaint.
+1 to the original post โ and I'll add a partner-side view, because I'm hitting the same thing while scaling certification across a team. The disjointedness you describe (videos โ slides โ notebook labs โ Vocareum/Uplimit labs, inconsistently) matches my experience exactly, and there's a worse operational version of it: content changes or disappears in place. Labs I completed and planned to route teammates through no longer exist at the same 'Learning plans' โ so a learning plan I hand someone can be actually a very different experience for them than it was for me.
For anyone managing more than one learner, that's the real blocker: you can't standardise a path that isn't stable.
A few specific quality problems I've run into, beyond the morphing content in place:
Inconsistent and outdated SDKs across modules. Different modules use different SDKs, and in some cases SDKs the vendor has already marked -legacy. A learner can't tell whether an inconsistency between two modules is a deliberate teaching choice or just un-updated content โ so you end up learning a pattern that's already deprecated, with nothing flagging it.
Missing structured pedagogy. The material tends to present concepts as flat lists โ "here's what we just told you" โ without the structures that actually turn information into understanding: comparison tables, decision matrices, structured summaries. The single most valuable additions would be things like a decision table/matrix for which client/API to use when (WorkspaceClient vs function client vs vector-search vs deployments), a model-selection decision matrix, and a serving-selection one (pay-per-token vs provisioned throughput vs external models). Right now those distinctions are scattered across prose and video; a learner has to reconstruct the structure themselves.
Some instructor-led (Uplimit) sessions added little. In more than one case, the "live" session was a person playing back a recording that's already sitting in the learning plan โ which doesn't justify the instructor-led format, and for a paying/partner audience reads as filler rather than teaching.
The partner discount-voucher process is applied conditionally, which doesn't match the official benefits. The Partner Program Benefits document lists the 50% certification voucher as an unconditional partner perk, but in practice it's being issued conditionally (tied to learning-plan-completion recency) โ a condition that isn't in the published terms. That inconsistency between the documented benefit and the delivered process makes it impossible to plan certification budget across a team.
And the voucher arrived so late it was useless to me. By the time it came through, I'd already had to sit (and pay full price for) the exam โ because the whole point of certifying is to test right after the training, while the memorisation and muscle-memory are fresh. Making the voucher land weeks after training completion defeats its purpose: you either wait and lose the retention, or proceed and lose the discount. The timing needs to align with when people actually take the exam.
Two things that would genuinely help:
- A single, current, canonical learning path per role, with old versions clearly retired rather than edited in place โ so a plan built today is still valid next month.
- Consistency and structure within a path โ one SDK generation across modules, and the decision-table / comparison-matrix scaffolding above.
Genuinely asked, not piling on: what's the right way to escalate this so it reaches the team that owns the learning paths (and partner benefits), rather than staying in the forum? Is there a product-feedback channel, an Academy/Learning contact, or someone who routes structural feedback to the content and partner-program owners?
Thanks for reading a long post โ several of us would adopt a current, consistent path immediately, and I'm genuinely happy to give detailed notes to whoever owns this. The platform is worth it, which is the whole reason I bothered to write.
Radoslaw Tomasz Gasiorek
