Careers at Bayesian Health: Teams, Pay and How to Get Hired

The Team You'd Join

Zero G Talent's data shows the 11-to-50-person team hires remotely across the United States and selects for engineers and scientists who can make probabilistic reasoning survive HL7 feeds, FHIR mismatches, and the particular chaos of Epic and Cerner deployments.

The live board shows six active postings, all remote-eligible in the U.S., Zero G Talent found: Software Engineer (Application Integration), Software Engineer (Data Integration), Software Engineer (Analytics), Infrastructure Engineer, Director of AI/ML, and Client Success Account Executive (Clinical). Four engineering tracks, one research leadership role, one clinical-facing commercial role. The split maps directly to the Intelligent Clinical Augmentation platform: ingest heterogeneous hospital data, normalize it, run inference at the point of care, surface the result inside existing workflows without alert fatigue. Integration roles outnumber the core ML role three to one.

The Director of AI/ML posting signals maturity. This is not a "build the first model" hire — it is a "scale and govern the model lifecycle" hire. Saria reported in a Day Zero interview the FDA's roster of 500-plus cleared AI tools and the need for predetermined change control protocols that let devices retune in the wild without re-clearance. The person who takes this role will own the statistical rigor behind those updates: defining priors, monitoring drift, designing the evidence thresholds that trigger a model refresh.

On the clinical-commercial side, the Client Success Account Executive (Clinical) role requires translating a posterior probability shift into language a chief medical officer trusts. The job description leans on the company mission — "empowering clinicians with the insights they need to make the right decision for the right patient at the point-of-care", and that phrasing is deliberate. Bayesian Health sells into health systems burned by black-box alerts; the hiring bar selects for people who have lived the workflow pain, not just sold into it.

Semi-annual destination trips and an equipment stipend are the only forced collision points for a fully distributed team. The interview loop tests whether a candidate can reason like the platform: update beliefs when new evidence arrives, resist confirmation bias, quantify uncertainty instead of hiding it.

Compensation

Bayesian Health pays like a well-funded startup that knows clinical AI talent is scarce. Base salaries start in the low six figures and scale sharply once you cross into staff and director territory.

Salary.com (July 2026) puts the company-wide average at $98,318, roughly $47 an hour, with a spread of $86,632 to $110,916 across individual contributors in engineering, data science, and clinical integration. The same source pegs a Senior Software Engineer at $111,230 average, bounded by $103,122 and $118,446. Treat these as market midpoints, not guarantees.

CandidatesReach slices by career stage. Early-career hires (mid-20s, associate or junior) see $101,000 to $122,000. Mid-level specialists and leads (early 30s) command $135,000 to $169,000. Senior staff and project directors (early 40s) land between $196,000 and $243,000. The mid-to-senior jump exceeds $60,000 — the shift from executing defined work to shaping clinical validation strategy, regulatory approach, and hospital deployment architecture.

Equity data is absent from public sources. Bayesian Health has not disclosed option pool percentages, strike prices, or refresh policies in any filing or job post reviewed. Candidates should ask directly during the offer stage; the company's investors include Andreessen Horowitz and Johns Hopkins.

Benefits follow the remote-first playbook: health, dental, vision, 401(k) match, home office stipend, paid holidays. The Work Index by Flexa notes the package includes health insurance, paid holidays, equity, and a remote workplace, plus semi-annual destination company trips and equipment support.

Bottom line: IC engineers and data scientists should expect $100K–$130K base. Staff and lead roles clear $160K. Director-level clears $200K. Equity is the variable that makes the total compensation conversation worth having — and the one you'll need to negotiate with data in hand.

The Interview Loop

Bayesian Health runs hiring through AshbyHQ; every recent role lists "Remote - US Only" as location. That remote-first posture shapes every stage, from scheduling to final debrief.

Publicly available interview data is thin: Glassdoor shows one interview question and two reviews posted anonymously. The company has not published its loop structure. What follows is inferred from the team's published work on probabilistic reasoning in clinical settings and from general remote-hiring practices at companies of this size.

The technical bar centers on probabilistic reasoning applied to clinical ground. Rather than generic algorithm puzzles, candidates should expect scenarios requiring Bayesian updating, for example, adjusting a sepsis risk score when new lab values arrive with known sensitivity and specificity. The prompt is deliberately underspecified: missing prevalence rates, ambiguous conditional independence assumptions. Interviewers evaluate whether the candidate identifies what they need, states assumptions explicitly, and computes a posterior they can defend. This mirrors the Bayesian model's function: quantifying and managing uncertainty by assigning probabilities to different outcomes, exactly what the clinical AI team does daily. Infrastructure and data integration candidates face a parallel track: designing a pipeline that propagates uncertainty from source EHRs through to the model endpoint.

Team interviews use a shared rubric. Interviewers score independently before a calibrated debrief, a practice that counteracts the halo effect by encouraging interviewers to evaluate each attribute independently. The debrief runs as a Bayesian update: each interviewer states their posterior probability of hire given their evidence, the group discusses divergences, and a final consensus probability determines the offer decision.

The process is designed as mutual inference. Candidates are encouraged to probe the team's own uncertainty: "What's the strongest evidence that would change your roadmap?" or "How do you handle label shift when a hospital system changes its coding practices?" Recruiters explicitly invite candidates to update their own priors about the role. The closing conversation includes budget transparency, and a specific thank-you email referencing a technical exchange from the loop is noted as a positive signal.

Offers go to candidates who demonstrate calibrated confidence: they state probabilities, not certainties; they distinguish aleatoric from epistemic uncertainty in their answers; they show they've already begun the clinical translation work in their head. The Director of AI/ML role requires evidence of shipping models clinicians actually trust — not just AUC improvements on held-out test sets. For individual contributor engineering roles, the bar is the ability to build infrastructure that makes probabilistic reasoning tractable at hospital scale.

Where the Work Gets Done

Bayesian Health operates from two physical offices and a distributed remote workforce. The registered headquarters sits at 666 Greenwich Street, Apartment 633, in New York City. A second office anchors in Baltimore, Maryland. LinkedIn lists both as primary locations.

New York functions as the commercial and regulatory nerve center. CEO Suchi Saria, named to TIME's 2026 TIME100 AI list, positioned the company to engage directly with hospital C-suites, CMS reimbursement pathways, and FDA review cycles. The May 2026 FDA clearance for its continuous sepsis monitor, and the CMS establishment of a first-of-its-kind Medicare reimbursement pathway for pre-suspicion sepsis monitoring effective October 1, 2026, were managed from this hub. The August 2026 addition of two physician executives, Dr. Martin Doerfler as Chief Medical Officer and Dr. Stephen C. Dorner as Chief Medical Officer for Care Transformation, signals that clinical strategy and payer-facing work run through New York.

Baltimore's role is structurally important. The city's density of academic medical centers, particularly those running Epic and Cerner at scale, creates a natural environment for the EHR integration work that defines Bayesian's engineering challenge: reading the full chart in real time, reasoning across longitudinal data, surfacing alerts inside the clinician's existing workflow without adding click burden. Published results from a five-site Nature Medicine study — 81% adoption, 89% adoption across customer sites, three-plus hours earlier lead time with 20x lower flag volume, the Day Zero interview reported, reflect validation work that typically requires deep, on-site collaboration with health system informatics teams. Baltimore provides geographic proximity to those partners.

Third-party aggregators list between 5 and 11 open remote positions depending on the source. For a company of this size, that ratio suggests remote roles now outnumber site-based ones. The shift aligns with technical demands: data integration engineers need access to hospital VPNs and FHIR sandboxes; infrastructure engineers build pipelines that ingest millions of clinical notes nightly; the Director of AI/ML leads a research agenda that publishes in Nature Medicine and presents at AMIA, conferences that are themselves hybrid.

What the physical offices enable is selective density. New York hosts leadership that must meet with hospital executives, FDA reviewers, and CMS officials in person. Baltimore hosts engineers and clinical informaticists who embed with health system IT teams during Epic upgrade cycles or go-live events. Everyone else, the core ML researchers, the backend engineers, the analytics team building dashboards that show trigger-to-consult conversion jumping from 10% to north of 40%, works from wherever they can maintain the compute access and data governance clearance the role requires. The company's own language, "real-time clinical intelligence that reads the full chart, reasons like a clinician, and surfaces the few patients who need attention directly in the EHR workflow", describes a product that lives inside hospital networks, not inside a headquarters. The offices exist to negotiate the contracts and relationships that let the remote team ship into those networks.

Who Lasts

The people who stay and advance at Bayesian Health share a specific intersection: technical depth, clinical curiosity, tolerance for the friction of deploying probabilistic systems inside regulated hospital workflows. The company's record, peer-reviewed publications in Nature Medicine, FDA breakthrough device designation and clearance, adoption across health systems including Cleveland Clinic and Mayo Clinic, selects for those who treat model performance as a starting point, not the finish line.

Probabilistic reasoning is the literal and cultural common denominator. The company's name references Bayesian inference; its core TREWScore system updates sepsis risk continuously as new chart data arrives, and the research consistently emphasizes calibration, lead time, and the tradeoff between sensitivity and alert fatigue. Candidates who think in terms of posterior updates, conditional independence, and decision-theoretic thresholds, rather than static accuracy metrics, map naturally to how the ML team frames problems. This role, listed as remote on the Zero G Talent board, explicitly sits at this intersection: owning model strategy for clinical decision support while coordinating with clinical validation and regulatory pathways.

Clinical fluency, or at minimum the humility to acquire it, is non-negotiable. The platform integrates with Epic and other EHRs, reads the full chart, notes, labs, vitals, orders, and surfaces risk in a way that mirrors clinical reasoning. Conrad Gleber, MD, MBA, FAMIA, Associate Chief Medical Information Officer at an early adopter site, said the system "fires 3 hours earlier and 20 times less often on patients who are actually getting sick. That's the difference between a tool that gets ignored and one that gets used." Engineers who have never shadowed a rapid response team or watched a nurse dismiss a pop-up alert will struggle to design for that reality. This role, also remote, requires translating between ML outputs and clinical workflows daily; the posting emphasizes deep understanding of hospital operations alongside technical credibility.

Regulatory and validation rigor filters for a specific mindset. The 18.2% sepsis mortality reduction reported in Nature Medicine came from a prospective, multi-site study across five academic and community hospitals over three years. The FDA clearance for continuous AI sepsis monitoring — the first of its kind, required evidence that alerts confirmed within three hours cut median time to antibiotics by 1.85 hours (95% CI 1.66–2.00). People who treat compliance as paperwork don't last. The Infrastructure Engineer and Software Engineer, Data Integration roles both touch systems that must satisfy FDA software-as-a-medical-device requirements; the team builds audit trails, versioned model deployments, and monitoring that satisfies both hospital IT security reviews and regulatory auditors.

Remote work amplifies the need for self-directed communication. The entire engineering and ML team operates remote or hybrid (U.S. only per current postings), coordinating with clinical partners on-site at hospitals. The Nature Medicine study noted that emergency department providers and those with prior alert interactions were more likely to engage; adoption is a human-machine teaming problem, not just a model problem. Bayesian's published "Adopt and Engage" behavior change model and clinician trust research make this explicit: the team designs for sustained use, not pilot enthusiasm. Engineers who default to async documentation, write decision records before code, and proactively schedule time with clinical end users thrive; those who need constant synchronous alignment stall.

Mission alignment acts as the final filter. Founder Suchi Saria started the company after losing a nephew to sepsis; the tagline "Now, when we say, 'we are doing everything we can' It is not a hope. It is a truth." appears on the company site. The work is narrow, covering sepsis, deterioration, pressure ulcers, and palliative care gaps, but the stakes are measured in mortality and length-of-stay reductions: a 5.5% absolute mortality drop in one deployment, 3.3% in another. Candidates who need broad consumer-scale impact or rapid iteration cycles without clinical review cycles will frustrate. The ones who stay treat each false negative as a patient harmed and each false positive as clinician trust eroded — and they build accordingly.

The platform updates sepsis risk with every new lab value. The hiring process updates its belief about a candidate with every answer. In both cases, the question is the same: what does the evidence actually support?

Working in AI? Zero G Talent tracks the openings: see every open Bayesian Health role, browse AI jobs, the companies hiring, and the people building the field.