Skip to content

Project Haven

Project Haven began as a research initiative at Johns Hopkins University, born from the intersection of biomedical engineering and clinical neurology, and became a medical-grade home monitoring system commercialized by Haven Health, Inc. under license from Hopkins. Samir Panda coded the project “Haven” during the grant proposal phase. The name remained because papers, institutional-review protocols, pilot families, and clinicians already used it before commercial release.

Haven’s significance extended beyond its technical capabilities. It represented the convergence of institutional research excellence and lived disability experience—a device shaped not only by engineers and neurologists but by a medical student named Logan Weston who understood what monitoring technology needed to do at 3 AM because he lived in a body that required monitoring and loved people whose bodies required more.

Development and Origin

The Problem

By the late 2020s, home monitoring for medically complex individuals remained dangerously fragmented. A family caring for someone with epilepsy, sleep apnea, and intellectual disability might use a baby monitor for video, a pulse oximeter clipped to a finger for blood oxygen, a seizure detection wristband for movement, and a CPAP machine with its own data stream—none of which communicated with each other. The caregiver had four different apps, three different alert tones, and no unified picture of what was happening. A drop in SpO2 combined with unusual movement combined with elevated heart rate meant something very different from any one of those signals alone, but no existing system could synthesize them.

The population most at risk of SUDEP (Sudden Unexpected Death in Epilepsy)—people with severe epilepsy and intellectual disabilities—was also the population least likely to tolerate wearable monitoring devices. Wristbands triggered sensory distress. Adhesive sensors caused skin irritation. Electrode-based systems required application by trained caregivers and were impractical for nightly use over months and years. The devices that could save lives were devices the people who needed them most couldn’t wear.

Project Inception (~2028–2029)

Project Haven originated as a collaboration between the Johns Hopkins Department of Biomedical Engineering and the Department of Neurology’s epilepsy division. The initial grant proposal focused on reducing SUDEP risk through continuous home monitoring that eliminated the need for body-worn sensors. The core research question was whether contactless sensing technologies—radar-based respiration and heart rate monitoring, AI-powered video movement analysis, and ambient environmental sensors—could be integrated into a single system capable of detecting seizure activity, respiratory events, and physiological deterioration during sleep with sufficient sensitivity and specificity to be clinically useful.

Samir named the project “Haven” during the proposal phase. The name carried his wish to provide safety and refuge when a loved one could not remain in the room. Its use in papers, institutional review protocols, and pilot households established it well before commercial release.

The project’s personal origin lay in the remote care of Samir’s father, Pradeep Panda, after a heart attack and later stroke. Samir’s mother, Sunita, watched Pradeep overnight, while Suchitra coordinated appointments, medication, household logistics, crises, and daily recovery in Odisha. Samir sent money and sought detailed updates from the United States but could not take a turn in the room. Haven addressed the practical problem of monitoring a loved one when another family member could not remain physically present. That family history complemented rather than replaced the project’s epilepsy and SUDEP research origin.

Logan Weston’s Involvement (~2029–2034)

Logan Weston entered Johns Hopkins School of Medicine in fall 2029 with a background in community health advocacy, early research experience in the Hopkins CRISPR lab, published work on diabetic neuropathy and post-trauma recovery, lived experience as a wheelchair user with chronic pain and traumatic brain injury, and years of practical familiarity with epilepsy through Jacob Keller.

Samir Panda, the postdoc who had conceived Haven and led its engineering and research team, recognized Logan’s unique position at the intersection of clinical medicine, disability experience, and caregiving knowledge, and invited him to join the project in an advisory capacity. The invitation was carefully framed: as much involvement as Logan wanted, with the explicit understanding that he could scale back at any time without consequence.

Logan’s initial hesitation was not insecurity. It was the specific wariness of a Black disabled man who had learned that being “the disabled voice” on a research project could become a trap—lived experience reduced to a line item on a grant application, consultation that amounted to a photo opportunity while the engineering team made the actual decisions. He had watched disability “consultation” happen at institutions that should have known better. He didn’t want to be the wheelchair in the brochure.

Samir’s framing—“the door is open and it stays open regardless of how far you walk through it”—told Logan this might be different. Charlie Rivera reinforced the message with characteristic directness, and Logan’s broader circle of friends and family echoed the sentiment: this was where he literally shone. The intersection of technology, patient care, and lived disability experience was not a niche Logan had been assigned. It was a space he had been building toward his entire life.

At his first meeting, Logan made the project’s contactless architecture a non-negotiable access requirement rather than a feature that could be displaced by a wearable fallback. Years of watching Jacob Keller struggle with sensory overload during seizure monitoring, learning Sofia Medina’s equipment and sensory preferences through Cisco, and using medical devices himself had shown him how often technically successful equipment failed in daily life. He also argued for tiered alerts that distinguished “check” from “run,” multiple linked room units, accessible controls, and privacy-preserving data routing. He returned for subsequent meetings and became a continuing clinical and lived-experience collaborator.

Over the following years of medical school, Logan’s involvement deepened organically. He contributed to clinical validation protocols, provided insights on what families actually needed versus what engineers assumed they needed, and advocated for design decisions that centered the caregiving experience—the person holding the monitor at 4 AM, not just the person sleeping in front of it. His understanding of alert tier design was particularly influential: he argued that existing monitoring systems failed because they treated all alerts equally, when the caregiver needed to know instantly whether a sound meant “check when you can” or “run.”

Clinical Validation and Consumer Release (~2031–2035)

The clinical validation phase involved pilot deployments in homes of families caring for people with epilepsy and intellectual disabilities—the population the system was designed to serve. Feedback from pilot families shaped iterative refinements: the camera’s infrared illumination was adjusted to avoid triggering photosensitive seizures, the radar sensor’s range was calibrated for rooms of varying sizes, and the alert system was redesigned multiple times based on caregiver feedback about what was useful versus what was noise.

Charlie Rivera and Jacob Keller both volunteered as testers of their own accord—not because Logan recruited them, not because the project needed subjects, but because they saw what the team was building and understood it mattered and decided their bodies could contribute. Charlie’s sleep profile provided an exceptional stress test for the system: his CFS/ME meant non-restorative sleep even when the architecture appeared normal, his POTS created autonomic instability visible in heart rate variability, his sleep apnea required clean CPAP data stream integration, and his EDS meant his body moved in ways that a standard movement-analysis AI might misread as pathological. If Haven could handle Charlie’s sleep, it could handle almost anyone’s. Jake’s combined generalized and focal epilepsy—including the myoclonic jerks that occurred in sleep-wake transitions and the tonic-clonics the system absolutely had to catch—combined with autism-related sensory sensitivities that had already ruled out wearable monitoring, made him exactly the user the contactless approach had been designed for. His partner Elliot Landry, who had served as Jake’s primary seizure first responder for years, suddenly had data: not “I think he had two seizures last night because I woke up twice” but timestamped respiratory and movement logs showing exactly what happened and when.

The consumer version of Haven reached the market around 2033–2034 through Haven Health, Inc., the company built to commercialize the university research while preserving its disability-informed design. Hopkins retained the underlying licensing relationship and research partnership. Haven was marketed not as a baby monitor or a general wellness device but as a medical-grade home monitoring system for people with complex health needs—a category that had not previously existed in consumer technology.

Specifications and Function

The Haven Home System consisted of several integrated components:

Contactless Radar Sensor: A millimeter-wave radar unit that read respiration rate, heart rate, and gross body movement through the air, without any sensors on the body. The radar operated at frequencies that penetrated bedding and clothing, detecting chest wall movement for respiration and micro-movements for cardiac rhythm. The contactless design was Haven’s core innovation—the feature that made the system usable for people with sensory sensitivities, intellectual disabilities, or motor impairments that made wearable devices impractical.

Infrared Camera with AI Movement Analysis: A low-light camera with machine learning algorithms trained to distinguish seizure activity (tonic-clonic movements, tonic stiffening, rhythmic myoclonic jerks) from normal sleep movements (position changes, restless leg movements, REM-associated twitching). The AI learned the individual user’s baseline movement patterns over a calibration period, flagging deviations from that person’s normal rather than comparing against population averages.

Ambient Environment Sensors: Room temperature, humidity, and air quality monitoring. In Sofia’s installation, those readings provided environmental context alongside respiratory, sleep, and movement data.

CPAP Integration: Haven could read the data stream from compatible CPAP and BiPAP machines, correlating apnea-hypopnea events with the system’s own respiratory and cardiac data. This integration meant caregivers could see not just that the CPAP was running but whether it was working—whether oxygen saturation was holding, whether the mask had been displaced, whether breathing patterns were normalizing.

Optional Pulse Oximeter: For users requiring higher-precision SpO2 monitoring than contactless methods could provide, Haven supported integration with a finger-clip or adhesive pulse oximeter. In the Medina household, this was used during Sofia’s illnesses when respiratory compromise was a concern, but not during routine nightly monitoring.

Caregiver Dashboard and Alert System: A unified app and optional dedicated display that synthesized all data streams into a single interface. The alert system used three tiers—green (all normal), amber (parameter deviation, check when convenient), and red (urgent, respond now)—with distinct audio tones for each tier. Alerts could be pushed simultaneously to multiple devices, allowing Claudia, Cisco, and Michelle to all receive notifications. Historical data was logged automatically, providing a record for medical appointments that replaced Claudia’s handwritten notebook for overnight monitoring (though Claudia kept the notebook anyway, because Claudia trusted her own handwriting more than she trusted any device).

Two-Way Audio: Allowed a caregiver to speak through the room unit, so that “aquí estoy, princessita” could travel from the kitchen to the guest room without anyone leaving the stove.

Interoperability and User Independence

Haven was designed to integrate with the technology ecosystems disabled people had already built for themselves—not to replace them, not to compete with them, and not to demand that anyone learn a new platform in order to stay alive overnight.

This interoperability reflected one of Logan Weston’s core design principles: disabled people already had their tech configurations. They had already set up their iPhones with VoiceOver or their Android with TalkBack or their smartwatches with health monitoring. A system that forced them into a proprietary ecosystem or required a specific device brand created more work for people who already carried too much. Haven needed to fit into the life the person already had, not demand they rebuild around it.

Data Standards and Health System Integration

Haven’s data architecture was built on HL7 FHIR (Fast Healthcare Interoperability Resources)—the same open standard used by major electronic health record systems including Epic, Cerner, and their successors. This was not a marketing decision. It was a clinical one. Logan insisted that Haven data be legible to the systems that already held the user’s medical record, because data that lived in a standalone app and couldn’t reach the user’s doctor was data that existed in a vacuum.

The FHIR foundation enabled several critical capabilities:

Import: Haven could pull relevant clinical context from participating health systems—medication lists, known conditions, seizure history, allergy alerts—so that the monitoring AI had context for what it was observing. A heart rate spike in a user taking a beta-blocker meant something different than a heart rate spike in a user who was not. The import was read-only, required explicit authorization from the user or an authorized legal representative when applicable, and updated periodically rather than in real time to respect both bandwidth and privacy.

Export: Haven’s logged data—sleep reports, seizure event timestamps, respiratory summaries, distress inference flags, CPAP compliance records, environmental trends—could be pushed electronically to participating health systems. For health systems running Epic, this meant Haven data could flow directly into the patient’s MyChart record and appear in the provider’s clinical dashboard as structured flowsheet data, available for the doctor to review before an appointment rather than relying on a caregiver’s verbal report or handwritten notebook. For systems running eClinicalWorks (ECW), athenahealth, or other FHIR-compliant platforms, the same data exchange was available through standardized APIs.

User-Controlled Sharing: All data sharing was opt-in, explicitly authorized, and revocable. The user controlled which health systems received Haven data and what categories were shared, using supported decision-making or an authorized legal representative when applicable. Access could be revoked through the Haven app. Logan insisted on category-level controls rather than a single “share everything” toggle because the person generating the data retained authority over where it went.

HIPAA Compliance: Haven was designed from inception as a HIPAA-compliant system. Logan’s involvement ensured this was not an afterthought bolted on during regulatory review but a foundational design constraint. All data was encrypted at rest and in transit. The system met the technical safeguard requirements of the HIPAA Security Rule, including access controls, audit logging, transmission security, and integrity controls. Cloud-stored data (for the historical logs and multi-device synchronization) resided on HIPAA-compliant infrastructure with signed Business Associate Agreements. The Haven team underwent HIPAA compliance review as part of the FDA clearance process, and Logan personally reviewed the privacy architecture with the same clinical precision he brought to everything else—because he understood that for disabled people, whose medical data was already used against them by insurers, employers, and institutions, privacy was not an abstract principle. It was protection.

TEFCA Participation: By the time Haven reached the consumer market, the system participated in the Trusted Exchange Framework and Common Agreement (TEFCA)—the nationwide health data exchange network that allowed standardized, secure data sharing across organizational boundaries. This meant a Haven user whose primary care doctor was in New York and whose neurologist was at Hopkins in Baltimore could have their sleep and seizure data available to both providers without manual transfer, faxing, or the user having to remember to bring a printout to each appointment.

Personal Device Integration

Haven paired with iOS and Android devices without requiring proprietary hardware. The caregiver app was accessible by design: screen readers (VoiceOver, TalkBack) could navigate every screen, voice control could operate the dashboard, and the interface followed WCAG accessibility guidelines not as a compliance checkbox but because the people most likely to use Haven—disabled people and their families—were also the people most likely to depend on accessibility features. Data synced bidirectionally with Apple Health and Google Health Connect, meaning a user who already tracked their health metrics on their phone could see Haven data alongside their step counts, medication logs, and heart rate trends in a single unified view. Wearable data from Apple Watch, Fitbit, or other compatible devices could supplement Haven’s contactless readings when the user chose to wear one.

User Independence and Data Routing

Critically, Haven did not default to caregiver mode. Each user chose how data and alerts were routed, with supported decision-making when needed. Sofia Medina shared alerts with Claudia, Cisco, and Michelle so that her chosen support circle could respond across both homes; the dashboard and its choices remained available to Sofia in accessible forms. Charlie Rivera routed data to himself and, with consent, to Logan Weston. Jacob Keller shared real-time seizure alerts with Jake, Elliot Landry, and Ava Keller.

The system adapted to the user rather than assuming a default. Some people kept their data private, while others chose family, partners, clinicians, or support workers to receive particular alerts. Haven treated those arrangements as individual consent decisions rather than rankings of independence.

Distress Inference: Pain Detection in Users with Limited Communication

A later feature addition—developed during the clinical validation phase and refined through post-release updates—gave Haven the ability to analyze heart rate patterns, movement changes, respiratory irregularities, and facial muscle tension (visible on infrared camera) to infer potential distress in sleeping users. The feature was designed specifically for users with limited verbal or motor communication—people who could not reliably articulate their own pain or discomfort, whether due to intellectual disability, nonverbal status, post-ictal confusion, or unconsciousness.

Logan Weston’s position on pain inference was characteristically precise and uncompromising: the system inferred potential distress. It did not diagnose pain. The distinction mattered. A heart rate spike during sleep could indicate pain, but it could also indicate a nightmare, a seizure aura, gastric reflux, anxiety, or a dozen other physiological events. The system provided data. The caregiver provided interpretation. And the user—if they could communicate in any modality, through any channel, by any means—always had the final word.

“The system is a bridge, not a replacement,” Logan stated during a clinical advisory meeting. “It tells you something might be wrong. Your job is to ask. And if they can’t answer, your job is to look harder—not to let the algorithm decide for them.”

The feature used a tiered confidence system: low-confidence distress indicators triggered an amber informational alert (“Haven detected elevated heart rate and restlessness—check when convenient”), while high-confidence indicators—sustained heart rate elevation combined with facial tension, respiratory changes, and movement patterns consistent with pain response—triggered a red alert. The system learned each user’s individual baseline over time, reducing false positives by flagging deviations from that specific person’s normal rather than comparing against population averages.

For Sofia, who communicated discomfort through speech, ear-pulling, gesture, expression, and AAC, the system supplemented existing communication. For a sleeping, unconscious, or post-ictal user, a physiological distress flag could provide an early signal before direct communication was available. In either case, the alert prompted attention rather than substituting an algorithmic conclusion for the person’s report.

The ethical tension around inferred distress was addressed directly in Haven’s documentation and in Logan’s published commentary on the system. The feature was opt-in, required the user’s explicit consent with supported decision-making when needed, or authorization from a legal representative when applicable, and required a clinical recommendation. Guidance emphasized that algorithmic inference never replaced direct observation, communication attempts, or clinical judgment. Logan insisted this language appear on the activation screen: “This tool supplements your attention. It does not replace it.”

The Physical Object

The Room Unit

The Haven room unit was a single integrated housing approximately six inches wide, four inches tall, and two inches deep—roughly the footprint of a hardcover book stood on its spine. The casing was soft-touch matte plastic in warm gray, with a slightly curved front face that softened the industrial geometry. It did not look clinical. It did not look like surveillance. It looked like something that belonged in a room where someone slept—closer to a high-end smart speaker than to anything with a medical supply catalog number.

The front face contained the infrared camera lens (a small dark circle, no larger than a pencil eraser, recessed slightly to avoid reflection), the radar sensor array (invisible behind the plastic housing—no external indicator of its presence), the ambient light and environmental sensors (a barely perceptible strip along the bottom edge), and a single LED status indicator: a small, soft-glow dot that pulsed slowly in muted teal when the system was active and monitoring. The teal was chosen deliberately during the design phase—warm enough to not read as clinical, dim enough to not disturb sleep, visible enough that a caregiver glancing into a dark room could confirm at a glance that the system was running. The LED dimmed automatically in response to room darkness and could be turned off entirely for users with light sensitivity.

The two-way audio speaker and microphone were integrated into the housing—no visible grille, just a series of micro-perforations in the casing that allowed sound through without breaking the visual surface. The speaker was engineered for vocal clarity at low volume: warm, not tinny, capable of carrying a human voice across a dark room without sounding like an intercom.

The unit mounted via a simple bracket system—two screws into a wall or shelf, the unit clicking into the bracket with a magnetic attachment that held firm but allowed easy removal for cleaning or relocation. It could also sit freestanding on a shelf or dresser top, weighted slightly at the base to prevent tipping. The power cord was USB-C, running to a standard wall adapter, with a six-foot cord length that gave flexibility in placement. The unit had no internal battery for continuous operation—it required wall power—but contained a small backup cell that maintained Wi-Fi connection and sent a “power lost” alert to all linked caregiver devices if the power was interrupted, so nobody woke up to a silent room and wondered if the system had been watching.

The unit came in two colorways: the standard warm gray and a softer warm white for rooms with lighter décor. Both were deliberately neutral—designed to disappear into a room rather than announce their presence. For families who had spent years surrounded by medical equipment that declared itself in beige plastic and flashing lights, Haven’s visual restraint was not an aesthetic choice. It was a philosophical one. The device served the person sleeping in the room. It did not need to remind them, or anyone else, that they were being monitored.

The Caregiver Display

The optional dedicated display was a seven-inch touchscreen tablet with the same soft-touch matte finish as the room unit, designed to sit on a kitchen counter, a coffee table, or a nightstand. It showed the room view in real time—infrared footage of the sleeping space, the CPAP unit visible if present, the form of the person under the blankets rendered in the gray-green tones of night vision. The vitals overlay—respiration rate, heart rate, room temperature, humidity—updated continuously in the lower corner of the display, small enough to be unobtrusive but readable at a glance.

The display’s screen brightness adjusted automatically to ambient light. In a dark bedroom at 3 AM, the screen dimmed to near-black, showing vitals in a whisper of light that wouldn’t wake a sleeping partner. In a lit kitchen during the day, it brightened to full readability. The touchscreen responded to standard gestures: tap for detail on any vital, swipe for historical data, long-press for settings. The interface was designed for one-handed operation and for tired hands—large touch targets, no small buttons, no gestures that required precision when precision was the first thing exhaustion took.

During Sofia’s carriage-house visits, the dedicated display lived in the kitchen where Claudia could see it while she cooked. The screen showed the guest-room sofa bed, CPAP unit, and monitoring data. Claudia, Cisco, and Michelle also used the app on their phones, and the same user profile remained available through the primary unit at Sofia and Claudia’s home.

Accessories

The Haven ecosystem included several optional accessories, available separately, that extended the system’s capabilities for users whose needs or preferences went beyond the base unit’s contactless monitoring.

Haven Band (Wearable Armband): A slim, flexible armband worn on the upper arm or wrist that provided continuous pulse oximetry (SpO2), heart rate, skin temperature, and electrodermal activity data—higher-precision biometric monitoring than the contactless radar could achieve alone. The band communicated wirelessly with the Haven room unit and integrated seamlessly into the same dashboard and alert system. It was designed for users who could tolerate a wearable: the band was thinner and lighter than most consumer fitness trackers, with a soft hypoallergenic silicone strap and no rigid housing pressing against the skin. The strap width was adjustable through interchangeable sizes, and the band’s outer face could be customized—color options included standard black, white, and gray, as well as a range of brighter options. Custom and third-party strap covers were available, because the design team recognized early that a device people had to wear every night needed to be something they didn’t resent wearing.

Sofia Medina’s Haven Band was pink and sparkly. Claudia offered the strap when the system was installed, and Sofia claimed it with the same certainty she brought to favorite jewelry. She wore it because it was pretty, it belonged to her, and it sparkled when the light caught it. The monitoring device had become an accessory she chose rather than an apparatus imposed on her.

Haven Pad (Under-Mattress Sensor): A thin, flexible pressure-sensing mat placed beneath the mattress that provided ballistocardiography (BCG) data—detecting heartbeat, respiration, and movement through mattress vibration without any contact with the sleeper’s body. The pad was designed for users who could not tolerate any wearable, including the band, and for whom the contactless radar alone didn’t provide sufficient sensitivity. The pad was particularly useful for detecting subtle seizure activity that the camera might miss in a heavily blanketed user, and for providing respiratory data in rooms where the radar sensor’s range or angle was suboptimal. The pad was waterproof, machine-washable (removed from the mattress for cleaning), and thin enough that the sleeper couldn’t feel it through the mattress.

Haven Clip (Portable Pulse Oximeter): A small finger-clip pulse oximeter that connected to the Haven system via Bluetooth for high-precision SpO2 monitoring during acute illness or respiratory concern. Unlike the Band, the Clip was not intended for nightly use; it was an escalation tool used when respiratory compromise was a concern. In the Medina household, the Clip came out during Sofia’s respiratory illnesses because sleep apnea and acute breathing changes made oxygen monitoring a clinical priority rather than a background data point.

Haven Go (Travel Unit): A compact, battery-powered version of the room unit designed for travel and temporary locations—hotel rooms, hospital stays, family visits, respite care facilities. The Go unit contained the radar sensor and environmental monitors but not the infrared camera (to simplify setup and reduce privacy concerns in unfamiliar environments). It paired with the same user profile and caregiver dashboard as the home units, maintaining continuity of monitoring and data logging across locations. Battery life was approximately twelve hours on a full charge—enough for a single overnight—with USB-C charging from a wall outlet or any standard external power bank, because hotel rooms didn’t always have conveniently placed outlets and hospital bedsides were already a tangle of cords. The Go unit was smaller than the home unit, roughly the size of a deck of cards, and came with a foldable stand and a travel case. For families who traveled with medically complex members, the Go unit meant the monitoring didn’t stop at the front door. Logan’s design principle applied here too: the system followed the person, not the other way around.

Haven Pack (Travel and Storage Bag): A purpose-built bag designed to carry the full Haven travel kit—Go unit, Band, Clip, power bank, charging cables, foldable stand—alongside the user’s other medical supplies. The Pack was not an afterthought accessory; it was designed from the ground up with input from caregivers and users who already traveled with medical equipment and knew exactly what failed about every other bag they’d tried. The main compartment was padded and organized with elastic-secured slots for each Haven component, so nothing rattled, nothing shifted, and everything could be located by touch in a dark hotel room. Additional zippered compartments held medication, CPAP supplies, extra straps, pulse oximeter, and whatever else the user’s medical kit required. The exterior was water-resistant, durable, and unremarkable—it looked like a day bag, not a medical supply carrier, because the people who used it had spent enough of their lives with equipment that announced their disability to strangers.

Critically, the Haven Pack was wheelchair-mountable. It attached to the back of a manual or power wheelchair via universal straps with quick-release buckles, sitting flush against the seatback without interfering with push handles or tilt mechanisms. This was not a feature request that came from a focus group. It came from Logan Weston, who had spent his adult life navigating the world in a wheelchair and who understood that a bag designed for disabled travelers that couldn’t be mounted on a wheelchair was a bag designed by people who had never met a disabled traveler. The mounting system was tested with multiple wheelchair models during the validation phase and refined until it worked on everything from lightweight sport chairs to heavy-duty power chairs with recline functions.

All accessories synced automatically with the Haven room unit and caregiver app. No separate setup was required—pairing was handled through the app, and data from accessories was integrated into the same unified dashboard alongside the contactless sensor data. Accessories could be added or removed at any time without recalibrating the base system.

The Sound

Haven’s default operating mode was silent. Unlike medical monitors that beeped continuously, Haven only produced sound during alerts. The three tiers had distinct tones designed to be distinguishable even through sleep: green acknowledgment (a soft chime, used only when manually requested), amber (a two-tone ascending note, meant to wake a light sleeper without triggering panic), and red (a sharp, insistent three-pulse tone that cut through anything).

The two-way audio introduced a different kind of sound: the human voice. When Claudia said “aquí estoy, princessita” through the Haven speaker, Sofia heard her mother’s voice in the room even though her mother was in the kitchen. The sound quality was warm enough—not tinny, not compressed—that it functioned as genuine vocal presence rather than intercom.

In the Medina Household

Haven was installed first in the home Sofia shared with Claudia around 2034–2035, replacing the consumer baby monitor previously used during naps and overnights. A second linked unit followed in the Medina Carriage House guest room within one week. Logan Weston recommended the two-unit system for Sofia’s epilepsy, sleep apnea, CPAP use, and changing support needs; Ezra Cruz funded both installations.

The transition from baby monitor to Haven was seamless for everyone except Claudia, who spent the first two weeks distrustful of any device that claimed to know more about her daughter’s sleep than she did. Michelle ran interference, walking Claudia through the app, showing her the display, demonstrating how the alert tiers worked. Cisco let his mother adjust at her own pace and kept his phone alerts set to maximum sensitivity regardless. Logan provided a single-page guide in Spanish—typed, not handwritten, because his handwriting was affected by his TBI—explaining what each alert meant and what to do. Claudia kept the guide on the refrigerator and eventually stopped needing it.

Multi-Unit Architecture: “It’s Not a Haven; It’s a Cage”

One of Logan Weston’s most consequential contributions to Haven’s development was his insistence that the system support multiple units linked to a single user profile. The engineering team had originally designed Haven for deployment in a single room—one unit, one location, one user. Logan’s response became one of the project’s defining quotes and a phrase that would later appear in the system’s own marketing materials: “If it only works in one room, it’s not a haven. It’s a cage.”

The insight came from lived experience on multiple fronts. Logan moved among his home with Charlie, the Band House, and Hopkins and understood what it meant to depend on equipment fixed to one location. He knew that Sofia Medina’s CPAP traveled with her every time she visited the carriage house because Claudia packed it with her medication and other overnight supplies. Monitoring that existed only in Sofia’s primary bedroom would disappear during family visits. Disabled people visited family, traveled, slept in guest rooms, and moved among the ordinary places of their lives. A system that could not follow them was tethering rather than access.

The multi-unit architecture that resulted allowed multiple Haven units to connect to a single user profile. Each unit calibrated independently to its room’s acoustics and dimensions, but all units fed into the same dashboard, alert routing, and historical data log. When Sofia was at the home she shared with Claudia, the primary unit was active. When she stayed at the carriage house, the guest-room unit was active. Sofia’s selected alert circle—Claudia, Cisco, and Michelle—could receive notifications regardless of which location she used.

In the Medina family’s case, two units were deployed: one in Claudia and Sofia’s apartment (the primary unit, where Sofia slept most nights) and one in the guest room of the Medina Carriage House. Both were funded by Ezra Cruz. Both were recommended by Logan. The installation at Claudia’s apartment happened first; the carriage house unit followed within the week.

In Daily Use

The room unit required little direct interaction in daily use. It remained quiet and unobtrusive while Sofia used her own communication devices, CPAP, routines, and chosen Haven Band as needed. She could participate in choices about alerts and monitoring without having to operate the room hardware each night.

Legacy and Lasting Impact

Project Haven represented a shift in how home monitoring technology was conceived—from fragmented single-function devices to integrated systems designed around the actual experience of caregiving. Logan Weston’s involvement ensured that the system’s design centered the people who would use it daily: not just the engineers who built it or the clinicians who prescribed it, but the mothers who checked it at 3 AM, the brothers who carried the phone with alerts set to maximum, the families who had been cobbling together inadequate monitoring solutions for decades because nothing better existed.

The project also represented a model for disability-informed research—not “about us without us” consultation, but genuine integration of lived experience into the engineering process. Logan’s contribution was not tokenistic. It was structural, shaping fundamental design decisions that made the system usable for the populations that needed it most.

For the Medina family, Haven was infrastructure. It let Claudia cook dinner while Sofia napped, told Cisco whether to check or to run, logged data for medical appointments, and carried Claudia’s voice from the kitchen when Sofia called after waking. It was practical, daily support built around Sofia’s movement between her own home and family visits.

Technology Medical Devices Samir Panda Logan Weston Sofia Medina Johns Hopkins