Blog
Engineering deep dives, product decisions, and field notes from building Tara and Sevah's staff-reviewed workflow for assisted living.
82 posts·4 categories·last updated Aug 23, 2026
Most families start pricing home care the week after something scares them. Here is how the money actually works — hourly agency care, informal family coverage, and what a monthly subscription covers instead — so you can compare options without guessing.
The senior population is growing faster than the workforce built to care for it at home. A detailed, practical walkthrough of what it actually takes to open a home care agency — licensing, structure, insurance, staffing, software, and marketing.
Companion sits with someone and keeps the conversation going. The device we're building next handles the physical chores around the house. Here's the plan, and what's real today.
For most of home care's history, the scheduled visit was the only moment anyone outside the house knew how someone was doing. What changes when that stops being true.
Most calls home default to medication and appointments because that's what feels urgent. A list of things worth asking about instead — the questions that keep someone a person first.
The adult child on the phone is usually the first to notice something's off — long before it shows up on a chart. What to do with that, and why it shouldn't have to stay that way.
Families researching care after a hospital stay usually find two paths: move, or manage alone. The option that actually fits most situations sits quietly in between, and almost nobody explains it clearly.
Decades of research out of Yale link how people feel about aging to how they actually age — memory, mobility, even lifespan. The story matters more than most families assume.
Blood pressure gets measured every visit. Loneliness gets a guess. At home, the gap between visits is exactly where connection quietly erodes — and exactly where no one is watching.
Family caregiving doesn't come with a bill, which is part of why its real cost is so easy to miss — lost hours, lost income, and the attention that never shows up on anyone's ledger.
The hardest part of caring for a parent from a distance isn't any single task. It's that the shift never technically ends. What that does to an adult child, and what actually helps.
"I'm fine" is the answer that ends a phone call, not the one that tells you anything. What actually changes in the weeks after a parent comes home, and why the people closest to it often see it last.
Whisper-family models were built to transcribe whole utterances. Force them to stream and you trade away the future context they need to be accurate. At the bedside, where the last word often changes the first, that tradeoff is sharper than it looks.
Task bots have a finish line — book the flight, set the timer. Open-ended companionship has no success signal and nothing to complete, which is exactly why it's so much harder to get right.
Bedside care doesn't yield to clever engineering alone. It needs people who've sat with the problem long before they knew it was one.
TTS leaderboards optimize for naturalness — how human it sounds. For a hard-of-hearing 88-year-old, the most natural voice is often the least intelligible. We tune for comprehension instead.
Drug names are long, rare, foreign-rooted, and full of near-homophones that general speech-to-text was never trained to tell apart. A single wrong med name in a SOAP note is not a typo — it's a patient-safety event.
The device sits with someone between visits. Every other name we tried described a machine doing a job. Companion describes the job.
Field notes from voice sessions with older adults. Competence and warmth in small doses, not a chatty friend or a robot.
Shadow-mode evaluation: a candidate provider or model processes real production conversations in parallel with the live one, and we compare outputs at scale — with zero patient impact, because only the incumbent ever reaches the speaker.
A bench is not a bedroom. Notes from putting CareOS hardware in real rooms: a backlight that became a nightlight, a speaker turned to mud, one bar of cell.
Conversational repair: detecting when Companion misheard a person and recovering gracefully, without falling into the 'sorry, can you say that again' death spiral.
TTS engines mangle names and drug names confidently and constantly. Custom lexicons and phoneme overrides are how Companion gets Mrs. Okafor — and metoprolol — right every time.
A founder note on the Sanskrit word sevā, what it means precisely, and the discipline a name like this enforces on the product.
A good family update is short, specific, and honest. Companion's daily summary tells you how the night went and what the nurse already did.
Bias does not stop at the microphone. When a model turns a conversation into a SOAP note, it makes editorial choices about what to foreground and what to soften — and the literature is clear about whose symptoms get downplayed. That is why a nurse reviews every note, and why every line is traceable.
WER is the metric everyone quotes and the one that predicts bedside quality the least. The numbers that actually correlate with a person being helped are entity error rate, latency p95, turn accuracy, and escalation recall.
Pendants and sensors call after the fall. Companion is built to notice the week before — the gait drift, the medication signs, the risky hour.
Calling a tool — record a symptom, page a nurse — in the middle of a spoken exchange, without the awkward silence that exposes the machinery underneath.
On-device TTS would mean zero latency and zero network dependence. It would also mean a robot voice in a hospital room. Why Companion synthesizes in the cloud, and what we keep on the chip anyway.
Multilingual people code-switch mid-sentence, and dementia often peels back the layers to a first language learned decades ago. Locking ASR to one language ID is how you build a device that slowly stops understanding someone.
A reminder to drink water every two hours becomes wallpaper. Companion picks the moment, the wording, and sometimes the offer.
A fall is the visible event. The small changes in how a person moves — days earlier — are the warning Companion is built to catch.
Open speech corpora are full of fluent younger adults reading clean sentences. They contain almost nothing that sounds like a 91-year-old with post-stroke speech at home with the TV on in the next room. You cannot model a distribution you never sampled — so we go and sample it.
Audio fixtures, golden transcripts, and tolerance thresholds for a voice pipeline that is flaky by nature — because ASR is non-deterministic and a strict golden test would fail on every green build.
A nurse charting from memory at 6pm loses most of what happened at 11am. Companion captures observations as they occur and drafts the note with provenance.
Elders process speech more slowly and pause longer mid-thought. Tuning Companion's pacing — silence windows, speaking rate, and prosody — is most of what makes it feel kind.
Companion has a camera and a microphone. By design, it keeps structured observations — not a feed. Here is the technical line and why we drew it.
Humans backchannel, overlap, and listen while they talk. Full-duplex turn-taking is where voice AI is headed — here's what the realtime providers hand us today and what stays genuinely hard.
Partial transcripts that keep rewriting themselves make a device feel uncertain and untrustworthy. How we kill the flicker — debouncing, a stability threshold of N unchanged updates, and committing only the prefix that has stopped moving.
Presbycusis eats consonants first. Our default voice settings are shaped around that fact, not around a generic listener.
Voice cloning is good enough now to recreate a family member from a few minutes of audio. For someone living with dementia, that is not a feature — it's a deception. Where we draw the line, and why.
Weak, breathy, hypophonic voices arrive at the bedside mic as low-SNR whispers. Crank the gain and you amplify the room; leave it low and you lose the person. The tradeoff has no free lunch.
A single aggregate WER can look healthy while the device quietly fails an entire cohort. The only honest way to find bias is to stop averaging and start slicing — by accent, age, gender, and clinical condition — and to read the worst slice, not the mean.
LLM-as-judge scales conversation evals to thousands of turns no human could read. It also brings length bias, self-preference, and confident wrongness — so the judge gets calibrated against nurses before we trust a single score.
Alert fatigue kills monitoring products before they earn trust. Companion is built on the opposite premise — silence is the default state.
Grounding and hallucination control when Companion is asked about medications and symptoms — and why the right answer is often a refusal that escalates to a human.
For the people we serve, the difference between a thinking pause and the end of a turn is the whole problem. Interrupting a word search is the cruelest failure Companion can commit — so we built around it.
Streaming partial transcripts let Companion react before the person finishes a sentence — but early partials are guesses the model will revise. How we read them for turn and intent signals without ever acting on text that isn't settled yet.
Default TTS prosody is tuned for fast, fluent listeners. Getting rate, pauses, and emphasis right for an 88-year-old means reaching for SSML — and discovering how unevenly providers support it.
AW88298 on TX, ES7210 on RX, one I2S bus, DMA in PSRAM. The path from datasheet to a conversation that sounds like a conversation.
Dentures, dry mouth, and missing teeth distort exactly the high-frequency sibilants and fricatives that speech-to-text relies on to tell words apart. It's a mundane, deeply eldercare-specific failure that general ASR was never tuned for.
Gender bias in speech tech is rarely loud. It shows up as a higher error rate on one pitch range, a synthesized voice everyone assumed should be female, and assumptions about who is speaking that the acoustics quietly encode.
Why Companion's voice loop on the CoreS3 is an explicit FSM with guarded transitions, and what that bought us in debugging and person trust.
Offline replay evals and online live metrics catch different bugs and miss different bugs. The trap is trusting either one alone — especially when realtime audio refuses to replay deterministically.
Multi-turn memory for one person, across a conversation and across days — what we keep verbatim, what we summarize, and what we deliberately throw away.
How CareOS resolves voice, prompt, and provider per-device at session start — Firestore for truth, Redis Ring for speed.
Acoustic endpointing hears a pause; semantic endpointing reads whether the sentence is actually complete. Companion fuses both, and the fusion is where the accuracy comes from.
A `context.Context` rooted at server startup lets us ship CareOS in the middle of the day without cutting a single bedside conversation short.
A small neural VAD is dramatically better than an energy gate at telling a whisper from a window unit. The question is not whether it is better — it is whether it fits in flash, RAM, and a 20ms frame on an ESP32-S3, and what we give up by running it server-side instead.
Speech tops out around 8 kHz of bandwidth. 16 kHz is the honest sample rate — anything higher just burns home wifi you don't have to spare.
Total synthesis time barely matters. Time-to-first-audio dominates whether Companion feels alive or broken. Why we chunk by sentence and start speaking before the reply is done.
Stuttering and disfluency don't just garble the transcript — the silent blocks and repeated sounds wreck turn timing, so a hands-free voice device interrupts the one person who most needs to be let finish.
Companion writes a crash record to NVS flash on its way down, then uploads it on the next boot. A nurse at 3am should never be a debugger.
Speech models are trained overwhelmingly on younger speakers. Aged voices — tremor, weaker pitch control, slower rate, more breathiness — fall outside that distribution, which means the demographic Companion exists for is the one the underlying tech serves worst.
Why CareOS uses a half-duplex gate instead of software AEC on Companion — and why turn-based feels right at the bedside.
Building a labeled conversation eval set from pilot data — consent, de-identification, and the uncomfortable truth that the cases worth grading are the rare, hard ones you have to go hunting for.
Why CareOS runs voice activity detection on the CoreS3 itself — and how a simple RMS gate keeps a bedside conversation feeling like a conversation.
Barge-in sounds simple — stop talking when the person starts. In practice it means cancelling a reply mid-word, flushing buffers across three machines, and not mistaking your own speaker for the person.
Why Companion fragments still frames across UDP datagrams — and why a dropped frame is, for ambient vision, the right tradeoff.
Endpointing — deciding the person has finished their turn — has exactly two ways to fail, and both are felt at the bedside. Here is how we frame the problem and what we measure.
Why CareOS uses a lock-free single-producer single-consumer ring between the mic task and the WebSocket pump — and what that means at the bedside.
A fixed energy threshold is the right place to start and the wrong place to stay. Why a single RMS number fails across rooms and people, and what spectral features and an adaptive noise floor buy us in the next iteration of on-device VAD.
Cloud realtime, neural vocoders, on-device synthesis — the TTS landscape is wide and none of it was built for someone's home. How we evaluate and choose per household.
Why every Companion waits a little longer between reconnect attempts — and why none of them ever give up.
Dysarthria — slurred, weak, imprecise speech from stroke, Parkinson's, or ALS — is common among people aging at home and brutal for speech-to-text. The acoustics are real signal; the model just wasn't trained to receive them.
Speech-to-text word error rates climb sharply for non-mainstream accents — and among immigrant elders aging at home, that gap compounds into a device that quietly stops working for the people who need it most.
Why Companion firmware speaks only OpenAI Realtime — and how the server quietly swaps the model behind it.
You can't tune what you can't measure, and at the bedside the thing you're tuning is a conversation with someone who may not be able to repeat themselves. So the harness comes first.
How we split network and audio across two cores so a 87-year-old talking to Companion never hears a crackle, stutter, or mid-word pause.
Where the milliseconds actually go between a person finishing a sentence and Companion starting to talk — and why crossing roughly 800ms makes the whole thing feel broken.
Why CareOS validates every device request with HMAC and a shared secret instead of a database round trip — and what that means at the bedside.