Virtual biotech operating model
A virtual biotech for evidence-led drug discovery.
Chief Orchestrator, specialist scientific agents and an independent Red Team connect opportunity selection to Candidate OS and MIDD — with humans retaining decision authority.
Operating boundary: Opportunity Scout is connected to a live read-only prioritisation backend. The Mission Room is still a guided simulation. Agent reasoning is advisory: scientific tools, evidence provenance, audit controls and qualified human review remain authoritative.
Surface disease, target and modality hypotheses before resources are committed.
Keep claims, sources, uncertainty and contradictions connected through the programme.
Red-Team reasoning tests assumptions, alternative explanations and failure modes.
Agents prepare decisions; qualified humans remain authoritative at consequential gates.
Opportunity, Discovery, Candidate OS and MIDD stay connected instead of becoming isolated analyses.
BayesArc capability map
One frontend over the systems we already have.
This view does not create a second scientific platform. It exposes the existing BayesArc, Agent OS, Discovery Twin, Candidate OS and MIDD capabilities as one company-level experience while the backend integration is completed later.
LIVE
Opportunity Scout
Evidence-weighted opportunity prioritisation with contradictions, unknowns, falsifiers and Red-Team context.
PRESENT
Agent OS + Discovery Programme Manager
Programme-scoped orchestration, specialist fan-out, durable runs, evidence packages and human approvals already exist in BayesArc.
PRESENT
Discovery Twin
Programme hypotheses, evidence claims, causal graphs, contradictions, credibility, experiments and governed decision state.
PRESENT
Discovery Red Team
Independent critique of specialist findings, opposing evidence, unresolved gaps and failure modes before progression.
PRESENT
BayesArc Discovery Execution
Target evidence, generation, docking, ADMET, developability, retrosynthesis and experimental handoff remain in the existing discovery stack.
PRESENT
Candidate OS
Candidate identity, evidence ledger, profile thresholds, decision lifecycle and governed nomination already exist.
PRESENT
MIDD
Programme-linked quantitative development, model orchestration and review-gated translational workflows.
FRONTEND NOW
Virtual Biotech Control Plane
Unified mission identity, company timeline and cross-system orchestration are represented here first. Backend wiring will follow without changing current client-facing APIs.
Frontend-only integration phase: this page currently presents the unified company experience and links to existing live BayesPharma capabilities. It does not create a new programme, change an existing client record, trigger scientific compute, or alter current backend workflows.
Organization
The virtual company hierarchy.
This page brings the existing leadership-agent idea and discovery scientist organization into one company-level operating model. The hierarchy is explicit so strategy, science, review, and authority do not collapse into one agent.
LEADERSHIP AGENT OS
Company orchestration layer
Leadership roles coordinate business, science, technology, governance, and model quality.
Chief OrchestratorSingle synthesis point for multi-agent company workstreams.
Agent CEOStrategy, priorities, partnerships, milestones.
Agent CTOArchitecture, APIs, reliability, security posture.
Agent CFORunway, pricing, cost and scenario discipline.
Drug Discoverer AgentTarget biology, docking, ADMET, synthesis, experiments.
Regulatory & QualityClaims, validation, readiness and review gates.
AI/ML EngineerModel evaluation, uncertainty, active learning.
Product SMEProduct boundaries and module fidelity.
Improvement LoopFeedback, traces, failures, and evaluation learning.
SCIENTIFIC COMPANY
Discovery scientist organization
The programme manager delegates a mission to bounded domain scientists and an independent red team.
Discovery Programme ManagerCreates the stage-aware plan, assigns specialists, and frames the decision gate.
Opportunity ScoutUnmet need, landscape and opportunity framing.
Disease Biology & Target EvidenceCausal rationale and biological evidence.
Medicinal ChemistryDesign, optimization and chemistry hypotheses.
DMPK & DevelopabilityExposure, ADME and developability liabilities.
Translational & Clinical EvidenceHuman relevance, biomarkers and clinical signals.
Portfolio & FeasibilityProgramme viability, trade-offs and resources.
Discovery Red TeamIndependent critique, contradictions and falsification.
Mission room
See how the company would work.
Preview how a company-level mission will coordinate existing BayesArc capabilities. This remains a browser-only frontend demonstration until the orchestration API is connected.
MISSION BUILDER
Frame one scientific objective
This frontend object previews the future company mission contract without writing to Agent OS, Discovery OS, Candidate OS or MIDD.
LIVE COMPANY FLOW
Project Aurora
Ready to simulate a governed workstream.
Virtual company frontend previewREADY
01
Chief OrchestratorClassify objective, risk, and required company lanes.
waiting02
Discovery Programme ManagerBuild the stage-aware scientific DAG and specialist assignments.
waiting03
Specialist scientist fan-outBiology, chemistry, DMPK, translation, portfolio, and evidence roles work in parallel.
waiting04
Discovery Red TeamChallenge claims, contradictions, assumptions and missing evidence.
waiting05
Discovery Twin / Scientific DAGConverge evidence into one traceable programme state.
waiting06
Human decision gateQualified human decides whether to advance, close gaps, redesign, pause or stop.
waitingOperating loop
Plan. Spawn. Execute. Challenge. Commit. Decide.
The company should scale through controlled decomposition rather than a single giant agent. Every temporary worker should have a bounded task, evidence contract, budget, and review destination.
Governed spine
One scientific state across the lifecycle.
The virtual company becomes useful when the swarm does not end in chat. Evidence needs to persist into the programme systems that already govern discovery and development.
Discovery Twin / Scientific DAGHypotheses, evidence dependencies, contradictions, uncertainty, experiments and decisions.
scientific state↓
BayesArcDiscovery programmes, therapeutic design, generation, docking, ADMET and experiment loops.
discovery↓
Candidate OSCandidate identity, evidence ledger, trade-offs, decision passport and nomination governance.
candidate↓
MIDDPopPK, PBPK/PBBM, PK/PD, QSP, dose strategy and trial simulation.
translation↓
Clinical DevelopmentTrials, biomarkers, endpoints, quantitative evidence and development decisions.
clinical↓
Regulatory / Safety / LifecycleSubmission evidence, benefit-risk, commitments, post-market learning and lifecycle change.
lifecycleScale roadmap
From virtual company to elastic scientific workforce.
The next step is not adding dozens of static agent names. It is building safe hierarchical fan-out and fan-in so temporary workers can process trials, papers, datasets, assays, molecules, and model checks at useful scale.
Unified Virtual Biotech mission contract
One company-level mission identity that references existing Opportunity Scout, Agent OS, Discovery Twin, Candidate OS and MIDD objects.
orchestration API laterCross-system orchestration
Coordinate existing Agent OS, Discovery OS and scientific services through one mission without changing their current client-facing contracts.
backend laterMass evidence fan-out
Trial-by-trial, paper-by-paper, dataset-by-dataset and candidate-by-candidate parallel workers.
elastic workersFan-in + governed commit
Deduplicate, challenge, rank and commit accepted evidence back into the authoritative programme spine.
review firstBayesPharma virtual biotech
Open current Agent OS
A company of agents — governed like a scientific organization.
This prototype deliberately separates the company concept from live backend execution. When connected later, the same interface can become the mission-control layer for bounded scientific swarms while Discovery Twin, Candidate OS, MIDD and human review remain authoritative.