Today, I was looking over lots of cool job opportunities, in particular in the life sciences space, and I was again struck by the fuzziness of the language we in the space use to describe our work:

I come from a manufacturing software background. In my role at Sepasoft, I managed MES (Manufacturing Execution Systems) software products that:

• track the component parts that make up a finished good,
• help plant workers understand and perform the task at their particular step or station,
• track KPIs like OEE (Overall Equipment Effectiveness),
• deep-dive into analyses and understand how well a particular run or shift went against expectations,
• and measure execution against work orders.

During the pandemic, I worked at the... Pandemic Response Lab, where I worked on a home-grown clinical LIMS (Laboratory Information Management System) called DaViD, a truly astounding backronym for Diagnosing Viral Disease. That solution:

• tracked the samples that made up plates, and the plates that got consolidated into further plates downstream,
• helped technologists and lab leads understand and perform the task at their particular step or station,
• track KPIs like TaT (turnaround time),
• deep-dive into a particular qPCR experiment, and understand which samples could be fully diagnosed, and which needed another attempt,
• and measure execution against the backlog of samples (as many as 10s of thousands a day during the Omicron surge!)

Then, at Neochromosome – a really cool biotech/DNA manufacturing startup – my amazing teammates and I built a workflow for a Next-Gen Sequencing lab in another home-grown LIMS called Rosalind (after Rosalind Franklin!). Rosalind, to this day:

• tracks the samples that make up plates, and the plates that get consolidated into further plates downstream,
• helps PhD scientists and research assistants understand and perform the task at their particular step or station,
• tracks KPIs like genetic coverage and the reads achieved by a given sample,
• deep-dives into a particular Illumina sequencer run, validate that samples hit coverage goals, and plan re-runs,
• and measures execution against large sequencing orders.

In this momentary, involuntary breather I'm in, it's interesting to zoom out and realize that notwithstanding how different all of this work felt in the moment, the deliverables in the end all ended up serving so similar a set of needs.

When PRL hired me, it was explicitly so that I would treat the process of making a LIMS in a very MES-y way – implement it such that while the traditional LIMS goals of sample tracking and validation would be achieved, the real goal was to ease the process of understanding the system's overall level of execution against our goals. This was reflected in the delivery of a very MES-ish dashboard we had up on all the screens during the huge Omicron surge – it showed the backlog of samples at each station, helped us to evaluate our bottlenecks, and showed us our throughput "scores" and turnaround times. Not the typical realm of a LIMS system!

It wasn't until late in my PRL term, during a lab leadership change, that I realized that what we had built was actually a LIMS/LES system. I learned this because the new leadership group had sight-unseen planned to be rid of our dear DaViD and replace it wholly with a COTS (commercial off-the-shelf) LIMS/LES solution. I was asked to present to them on what we had built, and it was really then that those seasoned hands explained that we hadn't really built a LIMS – LIMS solutions really tend to be more sample-oriented than the workflow-oriented tooling we had built. In particular, our qPCR analysis tooling, in which leads could view and analyze the Ct scores and curves produced by the qPCR reaction, was far beyond what any COTS solution provided out of the box. So, DaViD lived to fight another day, with the newfound understanding from these new leaders with regards to what we had created throughout the pandemic.

This experience, along with my experience doing it all again at Neochromosome, taught (or re-taught) that tooling is so secondary to what the tooling is for – you aren't "building a LIMS," or "building a LES," or "building a MES," you are empowering your org, whether clinical, biotech, or dog food, to see and understand the health and performance of the system, and where you are being held up. You aren't helping them figure out a series of edge cases, you are building a coherent, well-tracked, error-free workflow, and removing complexity.

When I've struggled, it's because I'm too in the weeds about a particular step or need.

When my teams succeeded, it's because we had that big, global picture in mind – it's not the tooling, it's the workflow.