From protocol to data, without a scoping call
The whole point is that you never wait on a human to tell you whether we can run your experiment or what it costs. You find out at submission.
Describe the experiment
Paste a protocol in plain language, upload a construct or sequence set, or POST a job from your pipeline. You do not have to learn a proprietary protocol language to get a quote.
Instant feasibility and price
The feasibility agent checks your protocol against what the fleet can currently execute and returns one of three answers immediately: runnable at this price, runnable with these substitutions, or out of scope with the reason. No quote cycle.
Ship samples or have them made
Send physical material, or have constructs synthesised and expressed on our side so nothing crosses a border that does not have to.
Autonomous execution
The protocol compiler turns your experiment into an executable instrument program. Runs are scheduled against the fleet, monitored live by the QC agent, and re-run automatically when a control fails rather than being delivered as a failure.
Structured data back
Results return as typed objects over the API alongside raw instrument files — each carrying a signed provenance record of instrument, reagent lot, operator programme, and parameters. Feed it straight back into the model that generated the hypothesis.
What you can order
Validation classes we build the fleet around — chosen because they are the steps that most often sit between an in-silico design and a defensible claim.
Construct & expression
Plasmid assembly, sequence verification, and protein expression — including cell-free synthesis, which removes the host-toxicity failure mode that kills cell-based expression of designed proteins.
Binding & affinity
SPR and BLI kinetics for designed binders. The single most common thing a protein design model needs checked.
Stability & biophysics
Thermal shift, aggregation, and structural characterisation for candidates that bind but will not survive formulation.
Functional & cellular
Enzyme activity, reporter assays, viability and cytotoxicity panels against relevant cell backgrounds.
Molecular readouts
qPCR, targeted expression panels, and sequencing readouts for perturbation experiments.
Clinical-context validation
Validation against Apollo-derived South Asian sample sets, where population relevance rather than raw throughput is the question. More →
FILL Publish per-assay pricing and turnaround SLAs here. Transparent published pricing is the core differentiator against both legacy CROs and Ginkgo — leaving it blank forfeits the advantage.
What you get back
- Typed results, not documents — every assay class has a stable schema, so a result is machine-readable the moment it lands
- Raw files preserved — instrument-native output is retained alongside the parsed result, so you can re-analyse without re-running
- Signed provenance — instrument identity, reagent lots, calibration state, and the exact executed programme travel with the result
- Negative results included — failed candidates are returned with the same fidelity as successful ones, because a design model learns as much from them
- Your data stays yours — results are not pooled into shared datasets without explicit, separately-consented opt-in