Planning Mimulus’s first DNA-storage service

Working with another member of the team, I sequenced the roadmap to prove value before building custom automation: start with a manual beta proof of concept storing and recovering data with standard lab steps, follow with a first multi-well card supporting individual file retrieval, and advance to a later card built for repeated reads while developing reader prototypes and software in parallel.

Starting from an inherited financial template that had zero cost of goods, I built our quarterly operating and unit-cost model around the real engineering work—outside hardware development, DNA synthesis, staffing, and realistic time to launch. When an early schedule assumed commercial sales right after funding closed, I reset the launch timeline around how long the card and surface-chemistry engineering would actually take. Once the team signed off on that financial model as our master record, I reconciled our development plan, hiring plan, and investor materials so every document told the same story.

When fundraising timing grew uncertain, I built a 10-month reduced-budget fallback plan that kept critical card, reader-breadboard, surface-chemistry, and encoding-software milestones moving for roughly a tenth of the funded budget while deferring full-time hires and a dedicated facility—and flagged an unfunded technical project-management role we would still need to integrate our suppliers’ work.

Early on, I also scoped a long-term concept for a self-contained data-center appliance combining DNA writing, storage, and reading—simplifying it around replaceable reagent cartridges and backward compatibility with our cards to keep the engineering credible before tabling it altogether so our budget and attention stayed on the nearer-term managed service.

The storage card, synthetic DNA, and reader all had to fit one another. With teammates working on the molecular biology, I evaluated competing DNA-synthesis technologies by what our service actually needed—how many unique sequences a chip could make at once, whether each spot yielded enough usable DNA to work with, and the cost per strand—showing why high spot density alone didn’t help if individual wells came up short on material.

Because dedicating an entire tiny storage well on a card to one small file would ruin the unit economics, I developed a multi-file addressing approach that let many files share a single well while still copying back only the requested file. I drafted and filed Mimulus’s first provisional patent application around the storage architecture and laid out a layered IP strategy separating our system-integration inventions from our partners’ chemistry and fabrication know-how.

For the beta cards, I split dry card manufacturing—where our microfluidics partner handled lamination design, fabrication, and well-recovery testing—from DNA loading, gating fabrication payments on design sign-off and tying later milestones to DNA receipt and loading.

Managing partners also meant owning bottlenecks on our side of the table. I shipped DNA test plates, flagged candidly when our own unconfirmed synthesis handoff spec had held up the partner’s loading test across three meetings, and caught where the partner’s proposed beta card storage workflow had drifted from our development plan before more investor materials went out. Our partners ran the physical bench work; I owned the requirements and interfaces needed to bring their work into one coherent system.