SaaS MVP development in San Diego.

The technical reality of shipping an MVP in San Diego.

Questions from San Diego founders.

We are building software for labs. What does the data model owe to scientific rigor?

Immutability, lineage, and units, before any feature. Lab software earns trust through properties consumer SaaS never thinks about: records are append-only with corrections layered as new versions rather than overwrites (a result, once recorded, is evidence); every derived value carries its lineage (which raw measurements, which protocol version, which instrument); units and significant figures are first-class types rather than display formatting; and sample identity is unambiguous across renames, splits, and re-runs. Building those properties into the Postgres schema from day one is cheap; retrofitting them after six months of casual updates is archaeology. The features that make demos sparkle (dashboards, collaboration) sit safely on top of a rigorous model and wobble forever on a casual one. We design the model with your scientists in week one, because they know where the bodies are buried in their current spreadsheets.

Our customers will include defense-adjacent companies. What does that mean for hosting and the MVP budget?

It means asking one qualifying question early: will the product store export-controlled technical data, or just ordinary business data from companies that happen to be defense contractors? The second case is common and underwhelming in its requirements: standard commercial hosting with strong security practices serves it, and contractors' procurement reviews are satisfied by the same artifacts other enterprises want. The first case is a different product tier: ITAR-relevant data demands US-person access controls and GovCloud-class hosting, and CUI brings CMMC expectations that cascade into your operations, not just your architecture. The honest MVP path when the answer is 'eventually the first case': launch on commercial hosting for the unrestricted market, architect the data layer so a partitioned gov deployment is an expansion rather than a rewrite, and price the gov tier as the separate project it is. Promising ITAR-readiness casually is how small vendors acquire unaffordable obligations.

Researchers are skeptical adopters. What earns a scientific user's trust in version one?

Reversibility, exports, and visible correctness, more than polish. Scientists adopt tools that respect their caution: every action undoable or at least inspectable (an audit trail readable by humans), data exportable completely and in formats that outlive your company (CSV, not just API promises), calculations shown rather than asserted (click a derived number, see its inputs and method), and nothing that silently modifies data (autosave yes, auto-correct never). Onboarding earns trust the same way: import their existing spreadsheet faithfully, including its quirks, and show a diff of what was interpreted. The anti-pattern is consumer-style magic: delight that hides mechanism reads as risk to this audience. Build the boring trust features first and the lab's most skeptical member becomes your champion, which in scientific sales is the entire game.

We will pilot with one research group first. How should the MVP be scoped for that?

Around that group's actual workflow, with the generalization deferred deliberately. The pilot-first scope: their sample types, their protocol structure, their instrument outputs (one or two file formats, parsed properly, not a universal importer), their reporting needs, and nothing speculative. This feels uncomfortably narrow to founders with platform visions, but it is the fastest path to the evidence that matters: a lab that runs on your product daily. The architecture holds the breadth: the data model uses general types (samples, protocols, results, attachments) even while the interfaces serve one group's specifics, so the second lab is configuration plus modest work rather than a fork. Budget-wise this keeps version one in the lower half of our range and spends the upper half only when a second, different group signs. Platforms in this market are earned one lab at a time.