Embryologists capture images in tight windows, often while juggling multiple tasks and working under lab constraints (gloves, variable lighting, intermittent network). Our previous capture flow made the job harder than it needed to be: too many steps, status that wasn’t always obvious, and quality checks that happened after the fact. The result was a noticeable number of retakes and a subset of images that weren’t consistently usable by downstream AI models. We also saw confusion around patient/sample matching when clinics were busy, and the UI didn’t adapt well to different hardware setups at the bench.

What we observed

• Retakes and “did this save?” uncertainty, especially at peak times

• Inconsistent focus/exposure/framing leading to lower “model-usable” rates

• Occasional sample mismatches when labels were similar or when staff were rotating

• Network hiccups causing silent failures; no clear local queue/retry path