Case studies
Three deployment patterns that show how the IP can be applied across CNS R&D, senior care and teleassistance — one voice-first stack, adapted to each operating context.
Longitudinal voice and conversation data built from daily senior interactions, structured for R&D teams studying aging, cognition and mental wellness.
Context
CNS research teams working on aging face the same structural problem: the populations they most need to study are the hardest to sample repeatedly. Prompted cognitive tests carry practice effects, test anxiety and patient burden, and quarterly site visits leave months of blind spots between data points. The partner needed a way to observe cognition, mood and voice as they actually evolve at home, without turning measurement into another clinical task.
What we deployed
Amigo runs daily conversations in the participant's natural environment and turns each call into a structured, versioned snapshot across cognitive, vocal, behavioral and mood dimensions. Raw acoustic and verbal markers are delivered as the base layer, with CogPath cognitive pattern analysis and VGDS mental wellness signal available as optional interpretation layers on top of the same stream. Every record ships with a data dictionary and audit lineage so the research team can trace how each value was produced.
Signal & outcomes
The result is daily, naturalistic sampling instead of sparse scheduled testing — hundreds of sessions per subject over months rather than tens at best. Intra-individual baselines make change detection personal to each participant, and the raw-plus-optional-layer structure lets the team validate their own models against the markers or consume CogPath and VGDS interpretations directly. The dataset is designed to fit existing CNS research workflows, with JSON, CSV and Parquet exports.
Amigo calls seniors, supports social and cognitive stimulation, and gives operators a continuous signal layer without adding screen complexity.
Context
A senior care operator wanted to support autonomy at home without asking residents to learn apps or manage screens. The people they serve are most reachable by voice, and the operator needed a way to maintain daily contact, encourage social and cognitive stimulation, and keep visibility on wellbeing between human touchpoints — all without adding operational or interface complexity for residents or staff.
What we deployed
Amigo provides a phone-first AI companion that calls seniors, holds natural conversation, and supports reminiscence, check-ins and daily engagement. The companion runs on CareLM, a GPT-OSS base model further trained for safe and adequate senior interaction under medical committee supervision, so pacing, repetition and distress are handled with senior-specific behavior rather than a generic assistant. The same conversations feed a continuous monitoring layer, giving the operator signal without a second data-collection workflow.
Signal & outcomes
Residents keep access through the channel they already use — voice, over the phone — while the operator gains an ongoing, longitudinal view of engagement and wellbeing. CareLM keeps interactions senior-safe by design, and because monitoring is a byproduct of companionship rather than a separate task, the signal layer runs continuously without added burden. The deployment can be hosted or delivered on-premise inside the operator's own infrastructure.
A custom voice agent receives or initiates calls, understands urgency, routes escalation, and keeps the conversation going when the person needs support.
Context
A teleassistance network handles a constant flow of IoT and assistance alerts, where the first minutes decide the right response. Not every alert is an emergency, and not every quiet alert is nothing — sometimes the person simply needs to talk. The network needed a voice layer that could respond immediately at any hour, understand what is actually happening, and route each situation without leaving anyone unattended.
What we deployed
A bespoke voice agent, built around the network's operations, initiates or receives calls the moment an alert fires. It opens the conversation, reads urgency from what the person says and how they say it, and triages the need — resolving low-stakes contact directly, escalating when the situation calls for it, and keeping the person engaged in between. The agent is designed to fit the network's existing escalation model rather than replace the humans in it.
Signal & outcomes
Alerts trigger an immediate, consistent first response instead of waiting in a queue. The agent detects and triages need in the moment, handing off to a human whenever a situation requires it, so staff attention lands where it matters most. When someone just needs support or company, the conversation continues — turning a raw alert into an appropriate response, whether that is escalation, reassurance, or simply being heard.
One stack
CNS data access, home autonomy and real-time alert response run on shared infrastructure — Amigo, CareLM, CogPath and VGDS — adapted to each operating context rather than rebuilt from zero.
Tell us the deployment context — senior care operations, a research program, or a teleassistance network — and we will map the right route on the Verbasync stack.