Self-designing labware for automated micro-plant phenotyping and sorting
Lab-in-the-loop platform that makes labware a tunable design variable: an AI agent designs and 3D-prints custom vessels for millimeter-scale plants, runs them on a robotic liquid-handling deck, and iterates from experimental data. Enables fully automated micro-plant phenotyping, miniaturized herbicide dose-response screening, and whole-plant sorting.
Overview
A lab-in-the-loop platform that makes labware a tunable design variable. An AI agent writes parametric OpenSCAD code, prints custom vessels for millimeter-scale plants such as duckweed, runs them on a robotic liquid-handling deck, and redesigns vessels from the resulting data. Because geometry is code, plate design becomes version-controlled and iterable in days.
The approach solves a common bottleneck in early-stage biology: standard (SBS) labware rarely exists for novel assays. AI-designed, printable consumables remove that constraint — vessels print at any site and can be injection-molded once a geometry converges. Applications include automated herbicide dose-response screening, micro-plant phenotyping, high-content imaging, and whole-plant sorting.
Technical specifications
Key features:
- AI-authored parametric CAD; the agent emits printable geometry and revises the next design and inspection protocol from results
- Deck-camera QC against printed fiducials for quantitative dimensional checks
- Robot-gripper-positioned microscope for in-situ high-content imaging of individual plants
- Duckweed chassis: clonal, 1–5 mm, 24–48 h doubling time
- Milliscale raceway ponds for optical selection and pipette-based sorting of single fronds
- Operates alongside liquid handlers, dispensers, and plate readers under a separate agentic control layer
Validation benchmark: automated miniaturized duckweed herbicide dose-response, scored against manual OECD 221 on EC50 agreement, coefficient of variation, plates per week, and hands-on time.
Technology readiness level
The system is already running: AI-designed printable labware is in active use on the automated deck. The 12-month validation plan is organized in four quarters, each ending on a testable deliverable:
- Months 1–3: material panel; watertightness and leachate toxicity screened with duckweed growth; fiducial-based dimensional QC; large-bore pipette planting
- Months 4–6: gripper-positioned microscopy; image-based frond growth analysis versus manual counts
- Months 7–9: miniaturized dose-response benchmark versus OECD 221 on reference compounds with known EC50s
- Months 10–12: peristaltic-pump flow cell for optical selection and sorting of individual fronds
The agent revises geometry from each result; iteration counts and measured improvements are reported.
About Colorado State University
Colorado State University is a comprehensive public land‑grant research university with an applied, partnership‑driven culture. Multiple research campuses—including the Fort Collins main campus, a public‑facing Denver site, and a foothills research complex with shared core facilities and pilot‑scale testbeds—enable companies to co‑locate, access instrumentation, and run validation studies. A statewide Extension network and proximity to the Front Range innovation corridor provide streamlined engagement with regional and national industry, while an integrated veterinary teaching hospital supports translational studies. Research is supported by competitive federal funding from agencies such as NSF, NIH, USDA, DOE, and DoD. A dedicated technology transfer office streamlines IP, contracting, and startup formation, with incubator and collaboration space for industry partners.
delivered to your inbox.