> A lot of the initial work is done as a "virtual" company with no lab. After that stage, if a true "wet lab" is necessary, one of the perceived barriers may b

> A lot of the initial work is done as a "virtual" company with no lab. After that stage, if a true "wet lab" is necessary, one of the perceived barriers may be, especially for the majority of modern software professionals, the situation with lab computing. Interpreting what former colleague (who is now a YC biotech founder) recently explained to me, labs often face significant challenges in needing on-premises computer hardware. [1] For some labs (DNA sequencing being a notable example, with the cost of sequencers plummeting), the sheer quantity of raw data flowing from the instruments can make (fully) cloud-based processing prohibitively expensive, if not impossible [2]. Having on-site servers hardware can be necessary, but, as a self-fulfilling prophecy, the knowledge about how to do this (supposedly) "nightmare" task well has been dying out. The far bigger issue, applicable to the vast majority of not all labs, is all the computers attached to the instruments and lab automation equipment, which are, in general, running proprietary software on a broad spectrum of Windows versions, which the users (lab) patch/maintain/administer at their peril (lest the proprietary software and, thereby, the instrument itself, stop working). Unfortunately (for me), the on-site server problem is generally much less dire and painful than the instrument-control problem. It can also be solved just barely well enough to cut down the size of the raw data, at which point software people breath a huge sigh of relief because they can ship it to the cloud and pretend hardware doesn't exist. The instrument-control/Windows problem ends up being hugely labor-intensive even to approximate a solution involving decent security and reliability. It also can require a remarkable amount of learning things that might never be used again in ones career, such as how a particular vendor's software communicates with a particular model and version of that vendor's instrument (so worth it, if it means making virtualization possible!). There are lab automation companies attempting to provide this sort of thing as a service, but they're not numerous, and it's unclear if this kind of IT is, so far, just a services/consulting add-on. [1] I consider my expertise in this area to be differentiating part of my Ops skillset, so it was of particular personal interest. [2] On a workable scale. With enough money, it should be possible to bring even 10-gigabit fiber Internet anywhere, or move the lab somewhere that already has it, but not in zero/negligible time.