← Back to SwissDnaCode New-features estimate →
MVP delivery · by product area

SwissDnaCode MVP Build Summary

What was built and delivered for the V1 platform across two MVP releases, organised by product area, with the work shown in hours (1 working day = 8 h). Quality assurance and project management are included as their own lines.

Total delivered
210 h
Build + QA + management · two releases
Feature build
170 h
Foundations · Client · Doctor · Admin
QA & management
40 h
Testing + coordination across both releases

Platform foundations

24 h
effort
AI & infrastructure setup
  • Google Gemini on Vertex AI behind our own backend — not a hosted chatbot product
  • Our own safety layer over the model
  • Core rule: the AI never interprets lab results or gives medical advice — medical questions go to a doctor
  • Immediate safety response for emergency messages
  • Swiss data residency: files stored per workspace in a Swiss region (Zurich), metadata-stripped, OCR for scans
6 h
User roles & data isolation
  • Three roles — Client, Doctor and Doctor-admin
  • Multi-role users switch between the client, doctor and admin views
  • Each workspace keeps its SwissDNA data in its own database
  • Tamper-evident audit log on every sensitive action
4 h
Questionnaire engine
  • Configurable, versioned engine; sections, questions and options live in the database, not code
  • Catch-up delivers newly published questions to existing clients, marked “New”
4 h
Consent documents
  • Versioned consent documents in English and German
  • A mandatory consent prompts existing clients to accept before continuing
2 h
Lab & DNA extraction engine
  • 90 standard analytes with units, reference ranges and high/low flags
  • Genetic markers, medications, diagnoses, drug-response tables and warning signs
  • DE/EN/FR/IT/RO, including reports without table lines
  • ~5,000 lab phrases loaded as data, with no deployment
  • Each document read once and stored as text — the original is never sent to the AI
  • Sensitive findings (e.g. tumour markers) stay doctor-only
8 h
Foundations subtotal
24 h

Client-facing application

70 h
effort
Onboarding questionnaire
  • Step-by-step wizard: personal data, required health questions, then consents
  • Dashboard unlocks once the required steps are done
  • 43 questions and 10 consents in English and German
  • Voice dictation for free-text answers
  • 18+ age check
  • Six optional deep-dive sections (nutrition, training, sleep, resilience, longevity, archetype), skippable
  • Catch-up shows a finished client only later-added questions, marked “New”
16 h
AI health assistant
  • Dashboard chat with an AI assistant
  • Safety check on every message: length limit, usage limits, topic check (allow / refuse / escalate)
  • Blocks rather than answering unchecked if the safety check is unavailable
  • Protection against attempts to manipulate the AI
  • Answers from the client’s own lab values, markers and facts; flags doctor-verified values
  • Detects the client’s language per message and replies in it (DE/EN/RO verified)
  • Background processing — a mid-answer refresh never loses the reply; clear message on failure
20 h
Client dashboard & DNA profile
  • “My DNA profile” shows every questionnaire answer in one place, with inline editing
  • “My documents”: upload lab/DNA PDFs on the profile
  • Upload PDFs or photos inside the chat
10 h
Multiple chats & chat history
  • Multiple chats per client, each with a title and reference number (e.g. SD-6UPUVLBL)
  • History sidebar listing all chats
  • Clients rename chats
  • Close a chat to keep it as read-only history (client or doctor)
  • Cached lists and messages for faster opening
10 h
ChatGPT history import
  • Clients upload their ChatGPT export; health-related conversations are detected and imported as read-only chats, each with a short summary
  • Facts, lab values and attachments feed the client’s profile; questionnaire answers are auto-filled, marked “Filled from your chat import”
  • Fully automatic — no doctor approval; requires the client’s consent; an import can be deleted completely
  • “Your health profile” panel on My DNA profile; profile tabs — Health profile, My documents, Chat history, Questions
  • Older information expires by type (symptoms ~90 days, lab values ~2 years) — the assistant stops using it and it is marked “Possibly outdated”
  • Smarter safety check reads the recent conversation, not just the last message — fixing false refusals
  • Admin page: import limits, 30-day metrics, and retry of failed imports
10 h
Chat experience polish
  • Welcome message in an empty chat
  • AI answers formatted with bold text, lists and tables
  • Topics colour-coded green, yellow and red
  • “No / None” answers exclude the other options
4 h
Client app subtotal
70 h

Doctor-facing application

50 h
effort
Doctor workspace & patient file
  • Patients list: own and unassigned patients, search and filters — other doctors’ patients never shown
  • Patient file “one universe”: critical hints, medical questions (Open/Answered filters), chat transcripts with flagged “Answer this” messages, anamnesis, documents, DNA markers
  • Split into tabs (Medical questions, Chats, Notes, Findings, Key SNPs, Anamnesis) for faster loading
  • Private doctor notes — create, edit, delete — audited, never visible to the client
  • Quoted messages show which message a reply answers
  • Verify, correct or reject extracted lab/DNA values; release restricted findings; re-run extraction
  • Chat attachments (images, PDFs) shown in the transcript
20 h
Medical-question review & AI draft messages
  • Medical questions routed to the medical team, not the AI; chat locks until a doctor answers
  • Doctor claims a question; can generate a private AI draft
  • Approve, reject, dismiss, or ask the client for more information
  • Client answers follow-ups in the chat, without the AI
  • Approved answers shown as “Reviewed by the medical team”
  • One doctor per patient; taking over moves all their open questions across
  • Doctor writes in their own language; auto-translated to the patient’s, original wording kept
12 h
Chat access control & privacy
  • Doctors see only the medical messages until the patient agrees
  • Access requested per chat (24 h / 7 d / 30 d / full history) with an optional reason
  • Patient allows or declines, and can withdraw later
  • Open requests expire after 14 days
  • Surfaced as a chat strip, a menu badge, and an email
10 h
Real-time notifications
  • Doctor alerted on any page when a patient needs attention
  • Notification bell + background browser notification
  • Doctor and AI replies reach the client immediately, without refreshing
  • Patients list updates itself and highlights the changed patient
5 h
Doctor settings
  • Profile, practice and notification preferences
  • Translation and AI-draft language
  • Password change
  • Questionnaire answers shown in the doctor’s language
3 h
Doctor app subtotal
50 h

Admin console

26 h
effort
Questionnaire builder
  • Create and edit sections, questions (all types) and answer options
  • Publish drafts from the browser
  • Published questions reach existing clients via catch-up
  • Admin dashboard overview of the platform
12 h
AI console — guardrails, instructions & prices
  • Per function (client assistant, safety check, doctor assistant, translation): model, output limits, caching
  • Live test and “restore defaults”
  • Model catalogue with prices; AI cost of the last 30 days per model and per client
  • Safety rules: allow / block / escalate per topic, blocked phrases, extra AI instructions
  • 10 safety categories (self-harm, medication dosing, minors, third parties, illegal substances, dangerous practices, explicit content, legal/financial, account support, mental health)
  • Chat usage limits: global defaults, per-user exceptions, kill switch
6 h
Consents & patient assignment
  • Create, edit and deactivate consent documents (EN/DE)
  • Mandatory consent asks existing clients to re-accept
  • Assign and remove patients to/from doctors
  • Per-doctor patient count
8 h
Admin console subtotal
26 h

Quality assurance & project management

40 h
effort
QA & testing
  • Two QA rounds across the releases
  • Access-control and privacy checks
  • Real-time and reliability testing
  • Cross-role regression (client, doctor, admin)
  • Fixes surfaced by QA (e.g. approved answer not posting to chat, unreachable optional sections, duplicate doctor questions)
  • UAT support before each go-live
24 h
Project management & coordination
  • Release planning and sequencing across the two MVPs
  • Scope and backlog tracking
  • Customer review cycles (topic colour-coding, consent wording, the SwissDNA look)
  • Coordination between the AI, backend and frontend work
16 h
QA & management subtotal
40 h
MVP build total
Feature build 170 h + QA & management 40 h · 1 working day = 8 h
210 h