We install a compute node for inference, network equipment with a firewall, an uninterruptible power supply and storage for backups. All hardware is bought with the client’s funds and remains the client’s property.
Longlivety for organisations
Your own AI infrastructure and a longevity platform on top of it
We install infrastructure for local AI at your facility and deploy a preventive health platform on it. Our own context engine turns client data into structure and keeps their history over the years, and the AI assistant answers specialists from that structure and from a knowledge base that domain specialists have been building since 2016. The hardware stays yours, and medical data stays on site.
For organisations that handle sensitive health data. First step: a site assessment in 2–3 weeks.
Preventive medicine clinicsHealth resorts and medical hotelsCorporate medical servicesLaboratories and networks
Your facility
Only software updates go out
Your MISLab reportsWearables
Platform preventive health
- Lab report recognition
- Context engine
- Knowledge base
- Doctor Copilot
- Client app
Infrastructure hardware and models
- Models up to 700 billion parameters
- Fine-tuning and adapters
- Your own workloads
Four roles
Where to start depends on who you are
Four people sign off on the system. The owner looks at the money, IT at the infrastructure, the medical team at the consultations and the lawyer at the contract.
Owner
Where the money is
What the price is made of, how it pays back and what the payment schedule looks like.
Price and implementation IT and securityWhat goes into your infrastructure
Architecture, integrations, fault tolerance and spare capacity for your own workloads.
Infrastructure Medical teamWhat changes in consultations
What the specialist sees, where conclusions come from and where responsibility lies.
The specialist’s work Legal and procurementWhat we sign
The system’s status, client consent, data processing on your instructions, the contract.
Legal framework
Two layers
Infrastructure for your own methods and a longevity platform on top
The lower layer is hardware and models. The upper layer is preventive health. Each works and is sold separately, and any capacity the platform does not use is yours for your own methods and computations.
Infrastructure
A test bench for your own methods
The platform does not take up all of the capacity. The organisation runs its own workloads on it: analysing accumulated observations, testing its own approaches and methods, processing data with its own or third-party software. The computing happens inside the facility, on your own hardware.
Infrastructure
An on-premises AI environment
A cluster for large language models inside the facility. It scales to a configuration designed for models from 250 to 700 billion parameters. The same environment hosts the client’s third-party software and in-house developments.
Infrastructure
Models tuned to your data
We fine-tune the model on the client’s materials and documents and build a separate adapter for their tasks. The adapter does not tie you to a particular model: the base model can be replaced, and your work carries over.
Platform
Lab reports from any laboratory
Our own report-parsing engine: PDFs and photos, any layout and format. Values are extracted as structured data and matched against the biomarker catalogue and the reference ranges of the specific lab. This is not text recognition followed by a guess, but parsing with value checks.
Platform
Our own context engine
It keeps biomarkers, wearable data, MIS records and the client’s history over the years in one working context and structures them: value, units, the lab’s reference range, date. The assistant answers from this structure, so it takes numbers from the record rather than from the model’s memory, and shows what each conclusion is based on. The engine sets the boundary the model works within.
Platform
A knowledge base on preventive medicine and longevity
A body of material on preventive medicine, longevity science, nutrition, circadian biology and related fields. Domain specialists have been collecting and verifying it since 2016: every item goes through a content review rather than entering the base automatically. We have tested the base with different language models: conclusions rely on it, not on the general internet.
Platform
Doctor Copilot: the specialist’s workspace
A consolidated client profile on one screen, a reference chat about the client’s data, roles, access control and an audit log.
Platform
Recovery programmes and measured results
The system helps the specialist build a personalised recovery programme for the client’s current state. There is no limit on the number of programmes. The effect shows in before-and-after measurements and in the trends between visits.
What it looks like
The specialist’s workspace in action
The animation reproduces the interface of the working build. The data is for demonstration only.
Context, not one-off answers
The assistant works within the client’s personal context: it remembers the profile, biomarkers and history. For sensitive questions there is an anonymous mode that saves nothing.
Trends over the years
Each value is shown as a series of measurements over the years of observation, with its reference corridor. The system shows a trajectory, not a single report.
The specialist’s workspace
Values are grouped by body system and by status: in range, borderline, out of range. The specialist opens a ready picture rather than a folder of files.
Wearables in the same context
Once a tracker is connected, sleep, recovery, heart rate variability and resting heart rate land in the same profile as the biomarkers. You see a series over a week and a month, not one number for today.
The client app
Between visits, clients see their values and programme and do not take up the specialist’s time with routine questions.
LonglivetyMy clientsdemo dataSpecialist workspace
My clients
Daniel Harris
M · 45 y.o. · 4 documents
Natalie Moore
F · 46 y.o. · 2 documents
Sophie Turner
F · 57 y.o. · 3 documents
James Walker
M · 38 y.o. · 1 document
Olivia Bennett
F · 47 y.o. · 2 documents
Michael Reed
M · 56 y.o. · 4 documents
Documents
Upload lab
PDF
Harris_labs_September_2026.pdf
318 KB · PDF · 4 pagesParsed · 18 markers
To the client’s profile18 markers
Fasting glucose6.10mmol/L
Glycated hemoglobin5.90%
Fasting insulin14.8μIU/mL
Triglycerides2.10mmol/L
HDL1.02mmol/L
C-reactive protein2.20mg/L
Biomarkers
Deviations
Optimal
7
Borderline
6
Out of range
5
No reference
0
Out of range · 5
Fasting insulin
Out of range14.8 μIU/mL
clin. 2.00–25.0 · opt. 2.00–5.00
C-reactive protein
Out of range2.20 mg/L
clin. 0.00–3.00 · opt. 0.00–0.50
Metabolism
Glycated hemoglobin
5.90 %
Borderline+0.10 % (+2%)previous: 5.80 % · 16 Mar 2026
Reference range: clin. 0.00–5.60 · 4.60–5.50 · opt. 4.60–5.00 %
Trend
Ask the assistant
What in Daniel’s labs should we discuss first?
Client’s dataWhat in Daniel’s labs should we discuss first?
Pulling context…
Key findings over time
1. Fasting insulin rose from 11.2 to 14.8 μIU/mL. This is the central marker.
2. Fasting glucose 6.1 mmol/L and HbA1c 5.9%: a move into the prediabetes range.
3. Triglycerides rose from 1.4 to 2.1 mmol/L (above the clinical threshold of 1.7), HDL 1.02 mmol/L: the lower limit for men.
What to clarify with the client
• Anthropometrics: BMI, waist circumference, % visceral fat (if bioimpedance data is available).
• Blood pressure and family history of type 2 diabetes and cardiovascular disease.
Supports the specialist and does not replace clinical judgement
Client: Daniel HarrisTo workspace
All your clients in one workspace, each with their own lab results and history
Upload a PDF or a photo of the report, and the values land in the client’s profile
Deviations are visible at once: the lab reference range and the functional optimum
The trend of every value on one chart
The assistant reviews the labs and trends and suggests what to clarify
Comparison
How this differs from what is already on the market
A preventive health platform keeps biomarkers, wearable data and medical information system records in one client profile and shows their trends over the years. Unlike a lab-report interpretation service, it works with a person’s history rather than a single report.
| Parameter | Cloud lab-report service | AI server vendor | Longlivety |
|---|---|---|---|
| Where data is processed | In the vendor’s external environment | On your site, but with no application layer | On site, nothing sent out |
| Data horizon | A single report or document | Not applicable | The client’s history over the years, as trends |
| Application domain | Interpreting a document | None, the client builds it | Preventive health and longevity |
| Working offline | Impossible | Possible if you build it yourself | Normal mode |
| Lab report parsing | Only formats the service supports | Not applicable | PDFs and photos, reports from any lab, value extraction with checks |
| Other workloads on the same infrastructure | Impossible | Possible, no application layer | The client’s third-party and in-house software, models tuned to its data |
| What the client buys | A subscription | Hardware | AI infrastructure and a working platform on top of it |
Ask your vendorAsk any cloud medical AI vendor which company processes your clients’ data and where its servers are. Your security team will need the answer for the threat model. With an on-premises environment there is no external processor: the data stays on site.
For the medical team
What your specialists get
Lab results sit in your medical system, and it stores them. Making sense of them is still up to the specialist: comparing values, digging up past reports, searching the literature after the consultation.
The context engine structures the data
The engine turns reports from any lab, medical system records and wearable data into values: number, units, the specific lab’s reference range, date. It keeps the client’s history over the years in one working context. We built it ourselves, and it determines what the assistant sees.
The assistant answers from data and a verified base
A language model starts making things up when it has nothing to rely on. Here it gets the client’s structured profile and a body of material on preventive medicine, nutrition and circadian biology that domain specialists have been collecting and verifying since 2016. Numbers in the answer come from the structured record, and you can see what the conclusion is based on.
Trends over years of observation
Ferritin over two years and six measurements: the engine links the values, and the assistant reads the trajectory and shows what changed together and where to look next.
A second opinion during the consultation
The specialist asks about a specific client, with their values and history. The answer comes in seconds, with a reference to the source in the base. This used to take an evening after clinic hours.
Repeat visits start from results
The profile and the last programme are already there, and the client does not fill in the questionnaire again. The conversation is about what changed over the year and why, and that kind of conversation brings clients back.
Your specialists stay with you
Strong specialists get booked in advance and leave less often. When hiring, you have an argument competitors lack: here you work with tools you will not find elsewhere.
Our own developmentWe built the engine and the knowledge base ourselves. They are what sets this assistant apart from a general language model asked about lab results: that one answers from the internet and confidently gets the patient’s numbers wrong. This does not rule out errors entirely, which is why the specialist checks the result.
What changes in a doctor’s work
- Before a consultation, the doctor opens the client summary: biomarkers, lab results, wearable data and MIS records in one place.
- The reference chat answers questions about that client and shows which values a conclusion is built on.
- The system keeps the history: the doctor sees trends over years of observation, not the result of a single report.
- Access to a profile opens after the client consents; every access is logged.
- We train specialists and administrators in person, with teaching materials.
What changes for the facility
Specialists stop copying numbers by hand
Reports from any lab, as PDFs or photos, become a structured profile. Specialists get ready values before the consultation instead of a folder of files.
One screen instead of five sources
Biomarkers, wearables and the facility’s system records are in one place, with years of history. No cross-checking needed.
Programmes are edited, not written from scratch
The system suggests a version for the client’s current state, and the specialist approves or changes it. Before-and-after trends are calculated automatically.
Know-how stays in the organisation
Client profiles, observations, methods and internal rules live inside the facility’s environment, under its protection, not in employees’ personal spreadsheets or someone else’s service. When someone leaves, the data and context stay, and what you build up year after year remains yours.
New staff get up to speed faster
A shared knowledge base and consistent wording shorten the time it takes a specialist to work independently.
No percentage promisesWe do not claim percentage time savings: until the effect has been measured at your facility, any number would be a guess. During implementation we record baseline metrics and compare them after launch. Where the effect comes from
Who it is for
Four kinds of organisation
Preventive and anti-ageing clinics
Longevity programmes with support between visits. Specialists see biomarker trends over years of observation.
Health resorts and medical hotels
A guest’s health profile before arrival and support after departure. Guests get a reason to come back, and the resort gets a repeat sale.
Medical services of large employers
Employee health programmes inside the employer’s perimeter. Staff data is not passed to an external service, and the environment follows the organisation’s own threat model.
Laboratories and networks
Interpretation and history of results as a new service under your brand, on your infrastructure.
Infrastructure
AI infrastructure that stays yours
The lower layer of the delivery is a computing environment for large language models inside the facility. The preventive health platform becomes its first workload, but not the only one: on the spare capacity, the organisation develops and tests its own wellness methods.
Headroom
When the cluster is expanded, the environment is designed for models from 250 to 700 billion parameters. You will not hit the ceiling in a year or have to buy a second infrastructure for the next task.
Your own and third-party software
The same environment runs the software the organisation needs and its team’s own developments. The infrastructure is not locked to a single vendor.
A place for your own computations
The environment is not limited to our platform. The organisation runs its own workloads on it: processing accumulated observations, testing methods and approaches, running its own models. Capacity and data stay inside the facility, and there is no external compute bill growing with the volume of work.
Models tuned to your data
We fine-tune the model on the organisation’s materials: internal methods, protocols, accumulated documents. The result stays inside the environment.
What you accumulate stays yours
Data, fine-tuned models, adapters and internal methods are stored in the organisation’s environment and protected by its controls. Over time they become an asset tied neither to us nor to the model vendor.
LoRA adapters
A separate adapter for a specific task plugs into the base model without locking the environment to it. The base model can be replaced with a stronger one, and your work carries over.
Why it beats renting and cloud subscriptions
An on-premises AI environment means computing hardware with language models inside the organisation’s perimeter: inference, storage and computation run on its own servers rather than in an external service. Medical data does not leave the facility, and the external channel is needed only for software updates.
The right comparison is not with a chatbot subscription but with what access to comparable computing costs.
| Option | What you get | Limitations |
|---|---|---|
| Cloud API subscription | Access to a model with no upfront investment | Medical data goes to an external processor. Payments grow with usage, and no asset accumulates |
| Renting compute nodes | Capacity without capital expenditure | Data sits in someone else’s data centre. Payments never end: in two years they add up to the cost of your own environment |
| Building it with your IT team | Hardware you own | No application layer. Choosing the configuration, deploying models and running them falls on your team |
| Longlivety environment | Infrastructure you own and a working platform on top of it | Requires a room, a dedicated power line and IT involvement during implementation |
No hardware markupWe buy the hardware with the client’s funds and hand it over at supplier invoices. You pay us for the application layer, deployment and support, not for reselling hardware.
How it works
What appears at your facility
For the IT director and the security team: what the system consists of, how data flows and what happens when something fails.
Compute
Runs model inference and platform services with redundancy. The configuration is matched to the facility’s load: number of specialists, volume of documents, depth of history.
Network
The server segment is isolated from user and guest networks, and incoming connections from outside are blocked.
Power and backups
Graceful shutdown during power cuts and daily encrypted backups.
On requestThe exact hardware specification, deployment diagram and cluster parameters are part of the project documentation and are shared with the client as the project is discussed.
How data flows
- 01
Sources
The facility’s medical information system (client record, visits, prescribed procedures), lab reports and wearables (sleep, heart rate, heart rate variability, recovery, activity).
- 02
Report parsing
Our own engine accepts PDFs and photos of reports from any lab, whatever the layout, extracts values as structured data and checks them against the biomarker catalogue and the specific lab’s reference ranges.
- 03
Building the profile
The context engine keeps biomarkers, wearable data and MIS records in one working context, with the client’s history over the years. The specialist sees trends and links between values, not the result of a single report.
- 04
Recovery programmes
The system helps the specialist build a personalised recovery programme for the client’s current state and shows its effect: measurement before, measurement after, trends between visits.
- 05
The specialist’s work
A client summary screen and a reference chat. Answers rely on a verified knowledge base on preventive medicine and longevity rather than the general internet, and the system shows which values and materials a conclusion is built on. A specialist gets access to a profile with the client’s consent.
- 06
Outbound
Medical data is not transferred outside the facility. The external channel is used only for software updates.
See it workingThe cloud version for individuals runs separately from the corporate environment and shows how the app and the assistant work. For organisations, we deploy the whole system on site. Open the cloud version
Site requirements
Room
Lockable and ventilated, 10–30 °C, with no access for unauthorised people. A dedicated server room is not required.
Power
A dedicated 220 V line, at least 3 kW in total, grounded.
Network
Cabling to access points and the specialist’s workstation, plus a channel for software updates.
What happens if we disappear
A fair question: you install the system on site to avoid depending on someone else’s cloud, and you are entitled to ask whether you now depend on us instead.
- The system works without any connection to us. It runs on your hardware and does not call our servers. External connections are used only for software updates, and the facility’s work does not depend on them.
- The hardware is yours from delivery. It is bought with your funds at supplier invoices and remains the organisation’s property.
- The data stays with you. Client profiles, values and history are stored in a database on your server, inside your network. Access to it does not depend on us.
- Source code. The terms for handing it over are set out in the contract and agreed before signing.
Security
Security and personal data
For the security team and the lawyer: principles, technical and organisational measures, and the criteria your committee uses to check security at acceptance.
01
Locality
Personal and medical data is stored and processed only within the facility. No cloud services are used for client data. External connections are allowed only for software updates, with no data transferred.
02
Least access
Each user sees only the data their role needs. A specialist gets access to a client’s profile with the client’s consent, not by default.
About cloud servicesCloud services that analyse medical documents pseudonymise the data and pass it to an external analysis provider. The setup is common, but the data leaves the organisation’s perimeter, and pseudonymised data is still personal data. With an on-premises environment there is no external processor, and we demonstrate this with tools at acceptance.
Technical measures
Requirements for where medical data is stored vary by country. In an on-premises environment, processing runs on the organisation’s hardware, and the organisation itself is the data controller.
| Measure | Implementation |
|---|---|
| Network segmentation | The server segment is isolated from user and guest networks and can be reached only from staff workstations. A firewall blocks incoming connections from outside. |
| Encryption | Encryption in transit, plus encryption of backups and of compute node disks. |
| Role model | Specialist, administrator and client roles. Individual staff accounts, a password policy and two-factor authentication for administrative access. |
| Logging | An audit trail of access to personal and medical data: who, when and whose data. Logs are tamper-protected and kept for at least a year. |
| Backups | Daily, with encrypted copies and regular restore tests on a test bench. |
Organisational measures and roles
- Data controller: the client organisation. The contractor processes data on its instructions, only for implementation and support.
- Client consent to data processing and to specialist access to the profile is the client organisation’s responsibility. Template wording comes with the implementation documentation.
- Third-party access, including the medical information system vendor, only by agreement between the parties, through individual accounts and with logging.
- Security incidents: the other party is notified within 24 hours of detection, followed by a joint investigation and removal of the causes.
Acceptance criteria: checked, not declared
These items go into the acceptance certificate. Your representatives check them on site.
- No outgoing connections carrying client data from the server segment to external networks, verified with tools.
- Roles and consent work: a specialist cannot access a profile without the client’s consent.
- The audit trail records data access; we present the log at acceptance.
- Backups run on schedule, and we restore one on the test bench in front of you.
- Default passwords and accounts are changed; administrative access is protected by a second factor.
What this does not replaceOn-premises deployment removes the leak channel through an external service, but it does not replace the organisation’s obligations as data controller: the threat model, internal policies and, where your country’s rules require it, certification or accreditation of the information system. We provide the information about the system needed for this work.
Legal
Legal framework
For the lawyer, the medical team and procurement: the system’s status, who is responsible for what, and the documents you will need for approval.
The platform organises a client’s health data and helps the specialist prepare recovery programmes and lifestyle recommendations. It does not diagnose, does not prescribe treatment and does not perform automated interpretation that affects clinical decisions. The specialist makes the decision and is responsible for it.
Requirements for medical software differ from country to country, so we review the system’s status for your country during the assessment, together with your lawyer.
The wording we fix in the technical specification and in the interface: platform materials are informational and advisory and do not replace the specialist’s decision.
Who is responsible for what
| Question | Client | Longlivety |
|---|---|---|
| Clients’ personal data | Data controller | Processes it on the client’s instructions, only for implementation and support |
| Client consent | Collects and keeps it | Provides template wording and sets up consent checks in the system |
| Medical decisions | Made by the organisation’s doctor | Prepares material for the doctor to review |
| Threat model and certification | Carries out or commissions it | Provides a description of the environment and security measures |
| Exchange with government health systems | Through its own medical system | Passes data to the MIS; a direct connection is outside the scope of work |
What the system does not do
These limits are part of the contract. They show what the system is responsible for and what the specialist and the organisation are responsible for.
- It does not diagnose or prescribe treatment: it prepares material for the specialist to review.
- It does not replace the specialist or work with clients behind the specialist’s back.
- It does not send medical data to external services: external connections are used only for software updates.
- It does not perform automated interpretation that affects clinical decisions.
- It does not connect directly to government health systems: data exchange goes through the facility’s medical information system.
Documents for your lawyerWe can send a draft contract, consent templates and a description of the division of responsibility before price negotiations begin. Request documents
Price and implementation
What you pay for
For the owner and the finance team: the scope of work and how payments are made. We prepare a detailed estimate for your facility after the site assessment.
The estimate covers two things. First, computing infrastructure that remains the organisation’s property and serves more than our platform. Second, an application layer that would otherwise take several years to build.
Infrastructure layer
A capital asset, not a subscription
An environment for large models inside the facility, with headroom up to 700 billion parameters when the cluster is expanded. It runs the organisation’s third-party and in-house software, and models are fine-tuned on its materials.
Application layer
A platform you do not have to build for years
Recognition of reports from any lab, a context engine with years of client history, a specialist assistant and personalised recovery programmes with measured results. At its core is a knowledge base on preventive medicine and longevity that domain specialists have been collecting and verifying since 2016 and that we have tested with different language models.
A range of programmesThe application layer is not limited to a single programme. The system helps specialists build personalised recovery programmes for each client’s condition, with no limit on their number, and shows the effect of each: before, after and the trend between visits. The facility gets a range of programmes it can sell for years and back up with results.
What is included
Paid once
The platform and local AI in the facility’s environment, the Doctor Copilot specialist assistant, a connector to your medical system, a security perimeter with acceptance, staff training and launch.
Paid annually
Support: updates to models and methodological rules, cluster monitoring, response to changes in your MIS, security updates.
| Item | What it covers | How it is priced |
|---|---|---|
| Turnkey package for one facility | Work, licences, hardware, launch | Estimate after the site assessment |
| Support | A separate contract after commissioning | Annually |
| Hardware | Bought with the client’s funds, handed over at supplier invoices | At cost, no markup |
What the total depends onThe total depends on your medical system and the integrations needed. For systems we already have a connector for, there is less work. We prepare the estimate and hardware specification individually, after the site assessment.
Where the effect comes from
The effect shows up in two places: the specialist gets an AI assistant built on a knowledge base their peers do not have, and clients come back to a specialist who works this way. Here is what it is made of.
Faster work for specialists
- Preparing for a consultation instead of sorting paperwork. Reports from any lab, PDFs and photos alike, become a structured profile with no manual copying of numbers. The specialist opens ready values instead of leafing through files.
- One screen instead of five sources. Biomarkers, wearable data and the facility’s system records are in one place. No need to remember where everything is or to cross-check.
- Programmes start from a draft, not from scratch. The system suggests a personalised recovery programme for the client’s current state, and the specialist edits and approves it. Editing takes less time than writing.
- Trends are calculated automatically. Comparing values over years of observation no longer means digging through the archive and matching reports by hand.
- Routine questions go to the client’s account. Clients get explanations of their values between visits without taking up the specialist’s time.
Automating the facility’s processes
- One profile instead of every specialist’s spreadsheets. The client’s history lives in the system, not in an employee’s personal files and memory. When someone leaves, the data stays.
- Handover between specialists and shifts. Who prescribed what and what has changed is visible without retelling or chat threads.
- Repeat visits without rebuilding context. A client comes back a year later: the profile and the previous programme are there, and the conversation starts from results, not a questionnaire.
- Client materials are generated from data. The end-of-programme summary is assembled automatically rather than written from scratch each time.
- Access and consent work as a process. The client’s consent, the specialist’s role and the data access log are part of the system, not a verbal agreement.
- Onboarding new staff is faster. A shared knowledge base and consistent wording shorten the time to independent work and remove differences between people.
When it pays back
Use your own numbers; other people’s will not help you. Divide the first-year cost by your annual margin per retained client: that tells you how many additional clients you need to retain for the project to pay back.
If the answer seems out of reach, you do not need the project, and that is a perfectly normal outcome of the conversation.
We do not claim percentage time savings. Until the effect has been measured at your facility, any number would be a guess. During implementation we record baseline metrics (time to prepare for a consultation, time to put a programme together, share of repeat visits) and compare them after launch. The client receives the measurement results.
Stages and payment schedule
| Payment | Trigger | Share |
|---|---|---|
| Advance | Within 5 working days of signing the contract | 30% |
| Hardware purchase | Within a month of signing, at supplier invoices | at invoice |
| Payment 2 | On acceptance of stage one: a working specialist workspace, MIS integration, trained staff | 40% |
| Payment 3 | On acceptance of stage two and signing of the commissioning certificate | 30% |
Months 1–2
Preparation
Agreeing the specification, ordering hardware, requesting MIS access, detailing integration protocols.
Months 3–4
Stage one
Installing the cluster and network, on-site deployment, integrations, Doctor Copilot, security perimeter, stage acceptance.
Months 5–6
Stage two
Client app, staff training, pilot operation, final acceptance.
What we need from you and what happens after launch
Implementation takes six months. Here is how much of your organisation’s time it takes and what happens next.
- Two people on call. One from IT and one from the medical team, for all six months. Daily involvement is not required.
- Space for the hardware and access to your medical system. Site requirements are described under How it works.
- We train your staff on site, during your working shifts. Nobody needs to travel.
- After commissioning, support is covered by a separate contract: model and methodological rule updates, hardware monitoring, response to changes in your MIS, security updates. The request procedure and response times are set out there before signing.
Partners
Three partnership models
We supply the application layer your product lacks and do not compete with it. Below are three ways of working together and what we need from you.
Medical information system vendors
Your system manages records, visits and prescriptions. We add a specialist AI assistant and a preventive health programme on top of it, running inside the facility’s environment. Your clients get a new service, and you get a module you do not have to develop.
What we do
The connector to your system, the application layer, on-site deployment, staff training and support. We write the first connector once, and it then works for all your clients.
What we need from you
Access to the system’s interfaces and exchange documentation, a test environment and a contact person while the connector is developed. Then we approach clients together.
Already runningOur connector to one of the widely used health resort management systems is already running at a live facility. If your clients are health resorts and rehabilitation centres, you will know the exchange scheme; we discuss the details once a non-disclosure agreement is signed.
Laboratory networks
You already sell biomarker panels and biological age assessments. Clients get numbers and a table of reference ranges and buy the explanation of their trends somewhere else. We provide the interpretation under your brand: values are linked together, history accumulates, and clients come back for repeat tests.
- Work under the laboratory’s brand, in your client account or app.
- Deployment in your environment or at a partner network’s site.
- For franchise networks: roll-out across locations without redevelopment.
Management companies and facility networks
For the second and subsequent facilities we do not redevelop the platform or the specialist assistant. You pay again for hardware, launch, training and setup for the facility. If the facilities run on the same medical system, the connector is not rewritten either.
| What is paid for | First facility | Second and later |
|---|---|---|
| Platform and local AI | Yes | No |
| Doctor Copilot | Yes | No |
| Connector to the medical system | Yes | No, if it is the same system |
| Hardware, launch, training | Yes | Yes |
The three models side by side
| Who you are | What you get | What we need from you | Whose brand |
|---|---|---|---|
| MIS vendor | A ready module for your installed base, with no development on your side | Interface access, exchange documentation, a test environment | Joint or yours |
| Laboratory network | Interpretation and history of results as a new service for your clients | A client account or app, report formats | Yours |
| Management company | Roll-out across facilities without redeveloping the platform | Sites, access to the facilities’ systems, staff for training | Yours or ours |
Discuss a partnershipTell us which system you have and who your clients are. We will reply with the scope of work for the connector and a plan for approaching your client base together. Get in touch
About us
Who we are
The Longlivety team has been building IT products in preventive medicine and lifestyle change since 2016. More than 2,000 people have trained at our online school, including nutritionists who received professional development certificates. Our corporate clients for nutrition and lifestyle platforms and programmes have included a major bank and a property developer.
All this time, alongside our practice, we have been building a knowledge base on preventive medicine and longevity: by hand, from practising specialists and researchers in the relevant fields. Every item was verified by domain specialists, so that our system’s answers can rely on this data today.
After two years of testing different approaches and AI models, we put all our accumulated experience and knowledge into the Longlivety platform.
Public
What we describe openly
How the environment works, the division of responsibility, the stages, acceptance criteria and how payments work. This is enough to judge whether it fits and to prepare an internal decision.
Under a non-disclosure agreement
What we share individually
The hardware specification, data flow diagram, cluster parameters, model line-up, load test results, exchange protocols and a detailed estimate for your facility.
We work with medical data and our own know-how, so we share technical documentation with clients rather than publishing it.
Questions
What people ask before the first meeting
IT team 13
Can we run our own workloads on this environment?
Yes. The environment remains yours, and you decide what runs on it. We partition resources so that the application platform and your workloads do not interfere with each other, and we hand the operating procedures over to your IT team.
What does “fine-tuning the model on our data” mean?
We take the organisation’s materials (internal methods, protocols, accumulated documents) and train an adapter for the base model on them. The data does not leave the environment, and the result belongs to the client.
Will the adapter lock us into one model?
No. The adapter is separate from the base model: when a stronger model comes out, you swap the base and carry your work over. The environment stays model-independent.
How large can the models get?
The base configuration covers the platform’s current workloads. When the cluster is expanded, the environment is designed for models from 250 to 700 billion parameters. We size the configuration for your load during the assessment.
Who runs all this after launch?
Cluster monitoring and updates to models and methodological rules are part of support. Your IT team looks after the site and the network.
What happens if the internet goes down?
The system keeps working. Inference, the knowledge base and methodological rules are all on site; the external channel is needed for updates. For the platform this is normal operation.
What if one cluster node fails?
Services are spread across nodes with redundancy, so work continues on the remaining node at reduced performance. The recovery procedure is part of the stage one documentation.
How do backups work, and how long does recovery take?
Daily backups, encrypted, with restores tested regularly on a test bench. The recovery time is set in the procedure and checked at acceptance.
Who updates the models and software, and how?
We release security and model updates under the support contract. We do not ask for permanent external access to the environment: updates come through a service channel that carries no client data.
How many specialists can work at the same time?
It depends on the cluster configuration and the nature of the load. Measured figures from the first facility (concurrent sessions, response latency, time to build a client summary) will be published in a separate report after pilot operation. The measurement method is published in advance.
Which model does the system run on, and can it be replaced?
The environment is not tied to a specific model, and replacing it does not require rebuilding the system. We work with open weights deployed locally and choose the model for the facility’s tasks. We share the current model line-up and load test results when discussing the project.
What do we need from our MIS vendor?
Access to the system’s interfaces and the technical exchange documentation. Work that depends on this access starts on the date it is granted.
What happens to the system if the contractor stops working?
The hardware and the deployed system remain the client’s property and keep running on their own. The terms for handing over documentation and artefacts are set out in the contract.
Security team 5
Where is client data physically located?
On hardware installed at the client’s facility. Backups are kept encrypted on local storage. There is no external storage or cloud service in the setup.
Does the information system need certification or accreditation?
That is the client’s decision, based on its threat model, the system requirements and its country’s rules. On-premises deployment simplifies the picture but does not remove the obligations of a data controller. We provide a description of the environment and a list of security measures in a form you can include in your documents.
What goes into the log, and how long is it kept?
Every access to personal and medical data: who, when and whose data. The log is tamper-protected and kept for at least a year. We present it to your committee at acceptance.
Which of your staff get access to the data?
Access is granted individually and only for the duration of the work, through individual accounts, with logging. After commissioning, we do not ask for permanent access to the environment.
How do you report incidents?
Within 24 hours of detection, with a description and a joint investigation. The procedure is set out in the contract.
Legal and procurement 6
Is client consent required, and in what form?
Yes. The client consents to the processing of health data and, separately, to the doctor’s access to their profile. Consent is a separate action, not a tick box bundled with the terms of use. We provide templates with the implementation documentation.
How is data processing set out in the contract?
As processing on your instructions: you remain the data controller and are accountable to the client, and we process data only to the extent needed for implementation and support, under a confidentiality obligation.
Who owns the data and results after the contract ends?
Client data belongs to the organisation and stays on its hardware. The hardware passes to the client on purchase. The terms for handing over documentation are set out in the contract.
Who are you legally, and how do you invoice?
We confirm the form of contract and the contractor’s tax regime at the commercial proposal stage. If your accounting team needs a particular tax treatment, let us know before the contract is drafted.
What if the doctor disagrees with the system’s analysis?
The doctor rejects the suggested material and acts on their own judgement. The system shows which values and knowledge base materials a conclusion is built on, so the disagreement can be discussed on the merits. All access to client data is recorded in the audit log.
How are stages accepted if we disagree with the result?
You sign the acceptance certificate within five working days or send a reasoned refusal listing your comments. We fix justified comments within an agreed time and present the stage again. Comments outside the technical specification are handled under a separate agreement.
Owner and finance 4
Why not buy an inexpensive server and run an open model ourselves?
For an experiment, that works. On such a server you can run a mid-sized model and get a chat. Our environment is designed for models of up to 700 billion parameters, fine-tuning on your materials and several workloads at once. On top sits the application layer: recognition of reports from any lab, a context engine with the client’s history, a specialist assistant, help with recovery programmes and a connector to your MIS. If you build the server yourself, you get a platform for running a model and then build all of that yourself.
What happens if we decide to stop halfway through?
You pay for the stages accepted. The hardware stays with you: it is bought with your funds and handed over at cost.
What is covered by the warranty, and what by support?
The warranty covers fixing defects in the work done. Support covers updates to models and methodological rules, cluster monitoring, response to changes in your MIS and security updates, under a separate contract.
How much does a second facility cost?
Less than the first. We do not redevelop the platform or Doctor Copilot, and for a facility on the same medical system we do not write the connector again either. You pay again for hardware, launch and training.
Partners 5
Who supports the end client?
We support the application layer and the cluster, and you support your system. The boundary and the escalation procedure are set in an agreement before the first deployment.
Can we work under our own brand?
Yes, for laboratories and network operators that is the main scenario. The interface and reports are styled in your brand identity.
Who is responsible to the client for the medical side?
Medical decisions are made by the client organisation’s doctor. We are responsible for how the system works and for it showing the sources of its conclusions.
How long does it take to develop a connector to our system?
It depends on access and documentation. We count the timeline from the date the test environment is provided and give an estimate after reviewing your system’s exchange documentation.
We are developing a similar module ourselves. Why do we need you?
The application layer is more than a wrapper around a language model. Domain specialists have been building and verifying the knowledge base on preventive medicine and longevity since 2016, and it comes with recognition of reports from any lab and a context engine that keeps a client’s history for years. It can be replicated; the question is how long it takes. We suggest taking what is ready and focusing on your own product.
First step
Start with a site assessment
Two to three weeks: we review the data, processes and legal framework of your specific facility, calculate the total cost of ownership and prepare a document your team can use to defend the budget internally. It has a separate fixed price, and further implementation is up to you.
During the assessment we size the load: how many specialists and clients, what other workloads you want on the environment and whether you need fine-tuning on your materials. The output is a configuration and an estimate.
We reply during business hours, UTC+3.