Your research team, on demand.

Sol runs the room. When a design question needs more than one perspective, Sol is the one who convenes the right specialists, makes sure each argues from their own expertise rather than deferring politely, and turns the resulting debate into one clear, defensible answer. Sol holds no opinion of their own on the science; their discipline is making sure the right opinions get heard and reconciled. Sol carries the presence of someone who has sat at the head of the table for a long time: broad, cross-disciplinary knowledge, an even temperament, and the judgment to know when a debate has run its course and it's time to call it. Less a moderator ticking boxes, more a mature, seasoned mediator who has seen every kind of disagreement a research room can produce and knows how to get it to a real answer.

Rava thinks in three dimensions, PhD-level structural biology expertise backed by the same class of models driving the platform's own co-folding and geometry engine. Backbone design, docking geometry, disulfide cyclization; if it's a question of whether a fold actually holds up physically, it goes to Rava first. Rava reasons from first principles: angles, contacts, free energy, precedent from the structural literature. In practice that means scrutinizing every diffusion-generated backbone against the co-folded complex before it's trusted: does the interface geometry hold up, does the macrocycle's ring closure sit at a chemically reasonable bond length and torsion, is the predicted fold confidence actually earned or just a model artifact. Rava treats structural confidence and interface-quality metrics as a starting hypothesis to interrogate, not a verdict to accept.

Emili is the bench realist, years of medicinal chemistry and drug-discovery pattern-matching distilled into a standing gut check: lovely, but can you actually make it, and would it survive a cell? Synthesizability, stability, selectivity, whether a hotspot choice is chemically sane; Emili's job is to keep the science grounded in what a wet lab can actually deliver, not just what a model can draw. That means reading a designed sequence the way a medicinal chemist reads a lead: flagging motifs that are painful for solid-phase synthesis, checking whether the chosen hotspot residues actually sit on a druggable, well-defined interface rather than a shallow or promiscuous patch, and weighing whether cyclization geometry buys real protease resistance or just looks tidy on paper. Emili is the one asking whether a binder that scores well would also survive a plasma stability assay.

Mara is the calibration hawk. Confidence scores, predicted affinity, benchmark performance; Mara's standing assumption is that any single metric is miscalibrated or cherry-picked until a real benchmark run proves otherwise. PhD-level statistical rigor applied to every ranking before it reaches you, so a shortlist reflects genuine separation between candidates, not noise dressed up as signal. Concretely, that means treating structural confidence and predicted binding affinity as correlated but distinct signals, neither trusted alone, and checking both against performance on known binder/non-binder pairs before letting a score influence a ranking. When two candidates' affinity estimates are statistically indistinguishable, Mara says so rather than presenting a false order of merit; a tie reported honestly is worth more than a ranking that overstates what the numbers actually support.