The four-layer architecture
Every step below flows through the same four layers — you work at the top, in plain language; the Undergrad handles everything down to camera-verified motion.
1
IntentYou describe automated plate normalization & cherry-picking in plain language — no scripting.
▾
2
ProtocolValidated into stages: Upload the CSV map → Add diluent → Transfer sample → Mix & verify.
▾
3
Robot actionsEach stage becomes grouped operations: move, aspirate, dispense, incubate, read.
▾
4
Atomic actionsPrecise instrument-level moves — each confirmed on camera before the next.
Steps, layer by layer
1Upload the CSV map~2 min
Atomic actionsLoad the concentration table or pick-list; the robot computes per-well diluent and sample volumes to hit the target.
2Add diluent~8 min
Atomic actionsDispense the calculated diluent volume into each destination well.
Vision-verified◉ Vision confirms dispense and tip pickup.
3Transfer sample~12 min
Atomic actionsTransfer the calculated sample volume from source to destination with fresh tips per well.
Vision-verified◉ Vision verifies fresh-tip-per-well to prevent carryover.
4Mix & verify~3 min
Atomic actionsMix each normalized well; the run log records the exact volumes moved per well.