Flagging What Actually Matters at Home
HealthcareAn anomaly-detection layer that learns each patient's own baseline and alerts a call centre only when behavior breaks from their personal normal.
Overview
A remote patient monitoring platform could tell when a sensor fired but not whether it mattered; a population-wide threshold was either too loose to catch anything or tight enough to bury the call centre in noise. Canvas Labs Partners originated, sold and delivered a behavioral anomaly-detection layer that learns each patient's own baseline and flags deviations from it, shipped to real patients.
Originated and sold pre-build
pitched, approved, SOW and payment terms locked before delivery began
Shipped to real patients
delivered through post-delivery testing within a single engagement
Tuned for sensitivity, by design
a high false-positive rate accepted on purpose given the cost of a missed emergency
Challenge
A house full of in-home sensors produced continuous data and almost no meaning: the platform could tell a carer that a pendant registered motion, not whether that mattered. In elderly in-home care, normal is personal rather than population-wide, so a single threshold across all patients was either too loose to catch anything or tight enough to bury a call centre in noise.
Solution
Canvas Labs Partners originated this project: pitched it, got it approved, then locked the SOW and payment terms before delivery began, on a fixed-fee agency engagement. The system learns each patient's behavioral baseline, sleep, bathroom routine, time home, time away, and a fall-detection pendant, then flags deviations from that individual's pattern rather than a population norm.
Alerts route to a hospital call centre, where an operator calls the patient, confirms what happened, and marks the alert right or wrong. That one decision keeps the system from making a clinical judgement, makes a false alarm cost one phone call rather than a clinician's time, and turns every confirmation into a labelled outcome at no extra cost.
The system was tuned for sensitivity rather than precision on purpose: a higher false-positive rate was accepted because a missed emergency can be fatal and a false alarm is only a phone call from an already-staffed centre. Canvas Labs Partners worked with engineering and design through the build, delivered into real patients' homes, and ran post-delivery testing.
Results & benefits
The project shipped to real patients: delivered, tested, and live within the time Canvas Labs Partners held the account. There are no accuracy figures, and that absence is a design position rather than a gap. Precision was deliberately traded for sensitivity, because a missed medical emergency can be fatal and a false alarm costs one phone call from a call centre already staffed to make them. The commercial side closed too: the statement of work and payment terms were locked before delivery began, on an engagement Canvas Labs Partners originated and sold rather than inherited as a brief.
Customer
Remote patient monitoring platform
Services
- Agentic AI consulting & development
- Applied ML & anomaly detection
- Healthcare product leadership
Tech stack
About us
Canvas Labs Partners is a software consulting team that designs and ships production agentic AI systems: multi-agent pipelines, tool-calling architectures, and human-governed automation, together with the trading infrastructure, blockchain, mobile, and product engineering those agents plug into. We publish our work as a case ledger rather than a portfolio, so every claim on this site links to the engagement behind it. Engagements follow the same four-step discipline, Map, Bound, Instrument and Govern, whatever the stack, and every entry in the ledger carries a result we are willing to be measured against.
Delivered individually, in a prior role, by one of Canvas Labs Partners’ founding team members. Not by the firm itself.
Have a system like this one?
We built this one end to end. If you are facing something similar in healthcare, book a free discovery call and we will tell you what it would take.