Most computer vision researchers don't dream about care worker timesheets. Ragav Sachdeva did, at least long enough to leave NVIDIA and Google's neural network labs for something closer to the ground: the administrative chaos that defines home care agencies.
His startup, TakeCareOS, emerged from Y Combinator recently with a premise that borders on the mundane—software to consolidate rostering, timesheets, care notes, and invoicing. Then it layers AI agents on top to handle what the company calls "the scut work." It's not the kind of moonshot that turns heads at tech conferences, but the timing might matter more than the technology.
Home care is big business, getting bigger, and increasingly scrutinized. Medicaid spending on home and community-based services in the US reached approximately $129 billion in 2022, up from around $115 billion in 2021, serving millions of users. Australia's National Disability Insurance Scheme has enrolled over 774,000 participants. Both systems are tightening compliance requirements. The 21st Century Cures Act's electronic visit verification mandate threatens funding penalties for states that fall short. Australia recently passed integrity legislation signaling a fraud crackdown and stricter oversight.
Which is to say: agencies need better tools, and they need them documented.
Six agencies, 200 employees, zero public names
TakeCareOS claims early traction—six agencies representing over 200 employees signed since launching pilots earlier this year. No customer names have surfaced publicly. The figures appear in the company's Y Combinator materials and Sachdeva's LinkedIn updates, where they're presented as proof of concept rather than market validation.
The agencies are running operations on a platform the company says can be deployed within two weeks. That's the pitch, anyway. The product targets disability, aged care, and home care providers juggling disparate tools for scheduling, documentation, and billing—specifically those navigating Medicaid HCBS rules in the US or NDIS compliance in Australia.
Atlas: the AI that drafts but doesn't decide
The software's AI component goes by Atlas. It functions as a conversational layer across the platform, drafting shift notes, suggesting compliance language, linking documentation to care plan goals, auto-converting PDF forms into digital entries. It reviews timesheet variances, completes incident reports, handles Medicaid invoicing workflows.
The company emphasizes restraint. "AI drafts and suggests — you review and approve," the website states. "Nothing is committed without your explicit sign-off." It's a human-in-the-loop design, meant to reassure operators in compliance-heavy environments where liability looms large.
Whether that balance satisfies agencies accustomed to manual control—or agencies desperate to offload administrative hours—remains an open question. TakeCareOS says it's working toward SOC 2 certification, though no completion timeline is public. The platform offers 256-bit encryption and data residency options for the US and Australia, baseline expectations in healthcare software but not exactly differentiators.
Beyond Atlas, the platform bundles standard operations tools: GPS clock-in/out with location alerts, credential expiry reminders, team messaging with channels and threads, recurring shift templates, audit trails, full-text search. It integrates with Xero, MYOB, and QuickBooks for financial reconciliation. Users access it as a progressive web app—installed directly from a browser, no app store required.
A crowded field of incumbents
TakeCareOS isn't exactly breaking into empty territory. WellSky rolled out multiple AI features recently, including ambient documentation for personal care and a triage tool called CareQueue. PointClickCare launched a next-generation EHR with AI-driven workflows for long-term post-acute care practice groups. MatrixCare holds significant market share in the LTPAC segment. In Australia, ShiftCare is a popular operations platform for NDIS and aged care providers, with its own AI capabilities.
The company's counter? It positions itself as an "AI-native operating system" rather than a traditional care management CRM. The framing emphasizes end-to-end workflow automation over point solutions stitched together. The company's blog includes competitive comparisons—one recent post breaks down TakeCareOS versus ShiftCare, with Atlas highlighted as the key differentiator.
How agencies perceive that distinction—whether it's a meaningful architectural advantage or marketing gloss—will determine whether TakeCareOS can pull users away from entrenched platforms.
From Oxford vision labs to care worker scheduling
Sachdeva's résumé doesn't exactly telegraph a pivot to healthcare operations. PhD from Oxford. Work in Andrew Zisserman's lab, a computer vision powerhouse. Internships at NVIDIA, Google, and Microsoft. Publications in computer vision spanning multiple years. It's the kind of CV that leads to research roles at AI labs, not SaaS tools for disability service providers.
The YC profile lists a team size of one, though that figure may lag actual headcount. The company hasn't announced external funding beyond Y Combinator's standard program support. Pricing isn't publicly disclosed—prospects have to contact sales. No customer testimonials or logos appear on the website yet, common enough for a company weeks past launch but notable in enterprise software where social proof drives adoption.
Compliance pressure as product tailwind
The regulatory environment in both target markets is tightening, no question. US states are phasing in EVV mandates with varying timelines and aggregator requirements. Australia's new integrity legislation raises the stakes for documentation and audit readiness. TakeCareOS has published guides on EVV compliance and Medicaid HCBS explainers, alongside NDIS-focused content referencing recent regulatory changes.
The winds are favorable, in other words. But favorable conditions don't guarantee execution.
Whether a solo founder—or a small, unlisted team—can scale a dual-market compliance platform against funded incumbents and deeply embedded workflows is the question that matters. The platform ships with a broad feature set, but enterprise healthcare software doesn't succeed on breadth alone. It lives or dies on implementation, support, and iterative refinement based on real-world feedback from agencies managing actual care workers in actual homes.
Six agencies is a proof of concept. The next year will clarify whether the AI agents deliver enough efficiency to justify migration costs—or whether long-term care operators do what risk-averse industries tend to do, which is stick with imperfect but familiar tools.
Sachdeva spent years training neural networks to recognize patterns in images. Now he's betting he can train agencies to trust software that recognizes patterns in compliance paperwork. It's a narrower problem, certainly less glamorous. But perhaps more lucrative, if he's right about the timing.
