Collaboration

The engine scales. Judgment doesn't.

Atlas Bio pairs agentic AI with Human-in-the-Loop scientific oversight. The engine handles the broad search; named experts own the irreplaceable judgment work. We are actively recruiting in two tracks — and the PatientPower portal is where we want you to start.

Two tracks

Where we need partners.

Track 1 · Start here

Practitioners for the PatientPower portal

Practicing clinicians who will use the patient-facing intelligence portal in their own clinic. This is the entry point — the fastest way to plug into Atlas Bio and the highest-leverage place for clinical judgment to shape the platform.

  • Who we’re looking for. Practicing MDs, DOs, NPs, PAs, and integrative practitioners across primary care, neurology, cardiology, endocrinology, oncology, and pain.
  • What you do. Use PatientPower with a small panel of consented patients. Flag where the platform’s recommendations match or diverge from your clinical judgment. Surface failure modes early.
  • What you get. Early access to the personalized intelligence engine, named co-authorship on validation reports, and a direct line into how the product evolves.
  • Time commitment. 2–4 hours / month. Quarterly review calls. Async feedback through the portal.
Apply for the PatientPower track →
Track 2 · Study reviewers

HITL reviewers for each study

Every Atlas Bio study — capsid, combo-safety, RWE, and the indication-specific pipelines — needs a Human-in-the-Loop reviewer with domain expertise. Reviewers see the pre-registered prediction, the audit trail, and the failure-mode flags before validation runs.

  • Who we’re looking for. Subject-matter experts in gene therapy / AAV, IO+TKI safety, neurodegeneration, cardiometabolic disease, oncology trial design, regulatory science, and clinical pharmacology.
  • What you do. One reviewer attached per study. Read the pre-registration, challenge the assumptions, and sign off on (or veto) the prediction before it’s SHA-256 locked.
  • What you get. Named on the manuscript and the pre-registration record. Reviewer access to the full evidence bundle and the IBC reasoning audit. Honorarium per completed study.
  • Time commitment. 3–6 hours per study, typically one study at a time. Asynchronous; no standing meetings.
Apply as a study reviewer →
Where to begin

Start with the PatientPower portal.

If you’re a practicing clinician and either track fits, start with PatientPower. It’s the most concrete way in — you’ll see the platform with real patients, in your own workflow, before deciding whether to take on a study reviewer seat.

Start with PatientPower →
How HITL works at Atlas Bio

Why a human reviewer is named on every study.

Atlas Bio runs a 19-agent IBC reasoning pipeline — Causal, Contradiction, Confidence, Mechanistic, Analogical, Abductive, Temporal, and Meta agents review and refine every prediction. The pipeline is rigorous, auditable, and pre-registered. It is also not a substitute for a clinician or domain scientist who has seen the failure mode in real practice.

The HITL reviewer is the named human accountable for the call. They see the audit trail before validation runs. They can veto, request a re-run, or flag a missing failure mode. The reviewer’s sign-off is what turns a model output into a defensible scientific claim.

This is how we caught our own hindsight contamination on the Alpha Predictor (96.6% → 91% honest). It is also how we keep the platform from drifting as new data flows in.

Ready to plug in?

Tell us which track fits, the indication or modality you work in, and how much time you can give. NDA available on request. Verified clinicians and sponsors unlock the manuscript bundle.

Contact Us →