The Handoff That Breaks Nurses
Kabilah, a Y Combinator-backed generative AI platform founded by two Stanford computer science graduates, has begun pilot deployments to replace paper nursing handoffs with what the company calls "cognitive infrastructure": a real-time digital report sheet that auto-populates from electronic health records and flags missing critical information during shift changes. The seed-stage company operates with a small founding team and a clinical advisory board drawn from Stanford Health Care, Advocate Health, Mount Sinai, HCA Healthcare, Banner Health, and Ohio State University. Its bet: the handoff is not a documentation task but a cognitive transfer, and the tools nurses use were never built for it.
Nearly half of nurses who quit name burnout as the primary cause, a figure Kabilah's website reports. The traditional response has been more training, more procedures, more forms. Kabilah's platform replaces paper report sheets with a Patient Advocate per patient, auto-filled from the EHR via FHIR, standardized for verbal handoff, and chat-enabled so the incoming clinician interrogates the patient's history instead of hunting through notes. The goal is not faster documentation. It is to make information usable in real time so the right context is understood and acted on at the right moment.
Sarah Raza and Umar Nadeem founded Kabilah in 2024 after converging on the problem from opposite angles. Both studied computer science with an AI focus at Stanford; both were Mayfield Fellows. Raza worked at Microsoft and EngFlow before co-founding AWARE. Nadeem spent time at Tesla. His motivation remains personal: a recent cancer diagnosis gave him an intimate view of the nursing workflows Kabilah now targets. "An intimate experience with the US healthcare system drives me to push positive change for the superheroes that saved my life," he wrote on the company's Y Combinator page. That stake, paired with Raza's systems background, shaped the thesis that generative AI can remove friction without adding another documentation layer.
Six People, Two Hubs, Advisors Who Open Doors
Raza, CEO, earned her BS and MS with that same focus, was the second hire at EngFlow, and founded a nationwide nonprofit. Nadeem, CTO, mirrors the Stanford CS-AI background, adds a master's in Management Science & Engineering, and joined the Mayfield Fellows program a year later. The company graduated from Y Combinator's Winter 2024 batch.
The clinical advisory board reads like a who's who of nursing informatics leadership. Gretchen Brown serves as Chief Nursing Information Officer at Stanford Health Care. Tracy Breece holds the VP of Nursing Innovation, AI & Emerging Technologies role at Advocate Health. Kenneth L. Davis, former CEO of Mount Sinai Health System, sits on the board alongside Ahmad Yousaf, Chief of Medicine at Saline Memorial Hospital with a Wharton MBA. Sherri Hess brings experience as CNIO at both HCA Healthcare and Banner Health. Michael Ackerman directs the Center for Healthcare Innovation and Leadership at Ohio State University. These advisors are not ceremonial; they shape product decisions and open pilot doors.
At AONL 2026 in Chicago, Kabilah hosted a Nurse Leader Listening Lunch facilitated by Ackerman. The LinkedIn recap from May 2026 captured three recurring themes: EHRs record what happened but don't help nurses do the work; tools aren't designed for how care is actually delivered; and any technology must preserve nursing instinct rather than replace it. The event wasn't a product demo; it was a working session with nurse leaders describing the gap between vendor promises and unit-level reality. That level of access indicates Kabilah is already embedded in purchasing conversations, not just pitching from the outside.
The Becker's Hospital Review feature published March 18, 2026, framed Kabilah as "cognitive infrastructure for nursing" rather than another documentation tool. Brown, quoted in the piece, said: "Documentation efficiency alone rarely moves quality metrics or length of stay at scale. Sustainable return comes when AI improves care delivery." That distinction matters. Hospital executives track length of stay, adverse events, and nurse turnover, not time saved on charting. A platform that claims to move those metrics needs clinical leadership willing to put their reputation behind it. The advisory roster suggests Kabilah has secured that commitment.
Two public testimonials name specific outcomes. Judy McLaughlin, a chief nursing officer, wrote that Kabilah "fits easily into our teams' day-to-day workflows and is helping us better serve our patients by eliminating miscommunication, reducing documentation burden, and ultimately decreasing burnout by keeping our nurses happier." Michael Scicchitano, a senior director of nursing quality and regulatory affairs, said the platform "aligns perfectly with our quality improvement initiatives by redesigning fragmented and complex workflows" and has "tangible impact on reducing adverse events through its intuitive data management capabilities." Neither testimonial identifies the health system, but the titles confirm deployment at organizations with formal quality and safety infrastructure.
Headcount figures vary. LinkedIn shows 11 to 50 employees and names four people: Raza, Yousaf, Josie Sutcliffe, and Allison Dempsey. The company's own site describes "a team of talented industry leading engineers, nursing leaders, and patient safety experts." The truth likely sits between the two counts: a core engineering nucleus plus clinical advisors who operate on contract or part-time arrangements.
Hiring signals where the technical bets live. Two founding engineer roles are open in New York:
| Role | Base Salary | Equity | Experience Required |
|---|---|---|---|
| Founding Engineer (Product) | $110K–$135K | 0.5–2% | 1+ years |
| Founding Engineer (AI) | $125K–$160K | 1–3% | 3+ years |
The gap in both compensation and experience threshold tells you the AI role is the harder fill and the higher leverage hire. Both roles demand presence in New York, suggesting the team is building a presence there or the founders have relocated to be closer to East Coast health systems.
The product-engineering split reflects a deliberate architecture choice. The product engineer will own the handoff interface, the surfaces nurses actually touch during shift change. The AI engineer owns the models that ingest messy clinical data and surface the right context at the right moment. That separation only works if the two sides communicate constantly. Kabilah's advisors, particularly Brown and Breece, have publicly emphasized that documentation efficiency alone doesn't move quality metrics; the AI must improve care delivery. That mandate lands squarely on the founding AI engineer.
Y Combinator's job board lists both roles without a closing date. The company's own YC page adds a direct ask: introductions to hospital executives and nursing leadership. The hiring push and the pilot push are the same motion. Each engineer hired expands the capacity to onboard the next health system. The team stays small by design until the product proves it can scale across divergent EHR environments and nursing cultures.
Why Pilots Stall at the Bedside
Kabilah's product architecture — FHIR-based EHR integration, real-time generative summarization, and per-patient "AI advocates" that chat against live clinical data — works in a controlled pilot. Making it reliable across dozens of health systems, each with its own EHR configuration, security posture, and nursing workflow, is a different engineering problem entirely.
The company says it can "go live in 2-4 weeks" and is "compatible with all major EHRs over standard FHIR protocols." In practice, FHIR implementations vary widely: Epic, Cerner, Meditech, and Athena each expose different resource profiles, extension patterns, and authentication flows. Kabilah's integration layer has to normalize those differences without breaking when a health system upgrades its EHR version or toggles a custom field. The platform "eliminates the need for redundant charting" by letting nurses "export their report sheets to various note types and flowsheets within the EHR." That write-back path — pushing structured, AI-generated content into vendor-specific note templates — is where most clinical AI pilots stall. A single malformed FHIR bundle can reject an entire shift's handoff data.
Then there's the generative layer itself. Each "Patient Advocate" must auto-fill customizable components from the EHR, generate a standardized verbal-handoff summary, and support interactive chat against the patient's full hospital course. Doing that in real time, at the bedside, with sub-second latency, means the inference pipeline has to run in US-based data centers (the company emphasizes data is "processed in highly secure, US based data centers and deleted after processing"). That rules out cheap GPU clouds abroad and forces a dedicated, HIPAA-compliant inference stack, likely self-hosted LLMs or a private VPC arrangement with a model provider. The platform also promises "all your data is encrypted in transit and at rest" and "customization to fit your security compliance and data governance needs, including Single Sign-On (SSO) for controlled user access." Every new customer adds a security review, a penetration test, and often a custom data-processing agreement. The engineering team has to bake that configurability into the deployment pipeline, not retrofit it per contract.
Reliability thresholds are higher than in consumer AI. The platform "guarantees every verbal handoff is thorough and accurate by notifying nurses of any missing critical information and providing corrections for inaccuracies." A hallucinated medication dose or a missed allergy isn't a bad user experience; it's a safety event. That means the summarization and correction models need rigorous evaluation against clinical ground truth, not just BLEU scores. Kabilah's "dedicated portal to monitor usage, including a multitude of key performance indicators and nursing quality metrics" is the instrumentation layer that makes that possible, but it also means the engineering team owns a feedback loop that spans product, clinical informatics, and hospital quality offices.
Scaling the human side is equally hard. The company promises "onsite support: our team of expert educators and engineers are available 24/7 to support on-site deployment." With a small team and two new health systems in the pipeline per the Y Combinator update, that commitment either burns out the founding engineers or forces the first dedicated hires into a hybrid role: half platform engineering, half clinical deployment. The company is also bringing its technology into nursing-school simulation labs, another environment with its own EHR sandboxes, faculty oversight, and student-data policies. Each new setting expands the matrix of integrations, security reviews, and clinical validation steps the small team must maintain.
None of these hurdles are unique to Kabilah, but the combination — real-time generative AI, bidirectional FHIR write-back, per-customer security hardening, and 24/7 clinical support — compounds fast. The next engineering hires will likely need FHIR integration depth, LLM ops experience in regulated environments, and a tolerance for hospital IT change-management cycles that move on quarterly, not sprint, cadences.
Not a Scribe. A Safety Net.
The nursing AI market is crowded. Directories such as Beyond the AI group Kabilah alongside tools like Colone, neuroClues, Docus, and Glow AI. But the founders and their clinical advisors argue the category itself has been misdefined. Most entrants treat nursing as a documentation problem; Kabilah treats it as a situational-awareness problem.
Kabilah's product scope reflects that framing. It targets every transition of care — emergency department admission, unit-to-unit transfer, shift handoff, discharge — and guarantees that each verbal handoff is checked for missing critical elements and corrected for inaccuracies in real time. Paper report sheets, still ubiquitous on medical-surgical floors, are replaced by a digital format that auto-populates from the EHR over standard FHIR interfaces and can go live in two to four weeks. A context-aware task list then prioritizes urgency across the nurse's assignment, while a chat interface lets the clinician interrogate that full course without navigating multiple charts. The platform integrates with existing tech stacks to eliminate redundant charting, and it ships with 24/7 onsite engineering and educator support, custom governance controls including SSO, and an analytics portal tracking nursing-specific quality metrics.
The competitive implication is structural. Tools that optimize note-taking after the fact — ambient scribes, summarization widgets, coding assistants — address a downstream symptom. Kabilah's bet is that the leverage point sits upstream: the bedside handoff where incomplete information propagates into medical errors, extended lengths of stay, and the redundant messaging that drives burnout. By designing for shift-based care, constant reprioritization, and team coordination (workflows the founders describe as "how nurses actually work"), the platform aims to preserve clinical instinct rather than replace it. That philosophy, validated by a Stanford Health Care CNIO and baked into deployment terms that put engineers on the unit floor, is the differentiator the company will test as it signs its next two health systems.
The FDA Question No One Has Answered
The FDA's oversight of AI-enabled clinical software creates the boundary conditions within which Kabilah must operate. The agency's device center has been explicit that its responsibilities are growing more complex due to advances in artificial intelligence, and its FY2022 budget request included $75.9 million for data modernization and enhanced technologies alongside $53.5 million for capacity building, signaling that the regulatory infrastructure itself is being rebuilt around the very class of products Kabilah is building. For a seed-stage company with a small team, the question is not whether FDA clearance will be required, but which pathway applies and how much evidence the agency will demand before a nursing handoff tool can sit on a hospital network.
Current FDA thinking on clinical decision support software draws a line between tools that merely display information and those that automate or drive clinical judgment. Kabilah's generative models, which summarize patient status and suggest handoff content, could fall on either side depending on how the product is positioned and how clinicians actually use it. The agency's Digital Health Center of Excellence has launched pilots such as TEMPO for digital health devices and a PreCheck program for domestic manufacturing, but neither has produced a settled framework for large-language-model outputs in acute care workflows. That uncertainty means Kabilah's engineering hires (particularly any future regulatory or quality lead) will need to design for evidentiary rigor from day one: audit trails for model outputs, version control tied to clinical validation datasets, and a post-market surveillance plan the FDA can inspect.
Expansion ambitions, as described in public materials, center on deepening pilot relationships with major health systems before pursuing broader deployment. Its site echoes that description, suggesting the founding duo intend to hire across clinical and technical domains in parallel. The competitive pressure is less about first-mover advantage and more about accumulating the real-world performance data that both hospital buyers and regulators will require. A clearance pathway (likely De Novo or 510(k) with predicate hunting) would unlock enterprise sales cycles that currently stall at pilot stage, but the timeline and cost are non-trivial for a team this size.
The hiring implication is clear: the next non-engineering role may not be a sales lead but a regulatory strategist who can map Kabilah's product architecture to the FDA's evolving expectations for AI/ML-based software as a medical device. Until that function exists in-house, the founding engineers carry the compliance burden themselves, which constrains how fast they can ship the core generative features that differentiate the platform. The same small team who must normalize FHIR across Epic and Cerner, harden inference for bedside latency, and staff 24/7 deployment support now also have to build the evidentiary record the FDA will someday demand. Nadeem's cancer treatment gave him the patient's view of the handoff gap; the next two years will test whether his team can close it at scale.
Working in frontier tech? Zero G Talent tracks the openings: see every open ASML role, browse frontier tech jobs, openings at Stripe, and the people building the field.