Voice and text logging
Freeform entries are structured into meals, symptoms and timestamps, then shown for approval before they become part of the diary.
Independent product case study
Turning an Abandoned Food Diary Into a Usable Product
Find My Triggers is an AI-assisted food and symptom tracker I created after conventional diaries kept becoming too much work. It combines fast logging with an analysis policy designed to surface uncertainty, helping people organize possible associations without pretending noisy health data proves a cause.
- Product
- Find My Triggers
- My role
- Founder & Full-Stack Product Builder
- System
- SvelteKit PWA with SQLite
The problem
The useful data only appears if logging is easy enough to continue
Food and symptom diaries usually ask for too many fields while giving little value back. Delayed symptoms, mixed meals and changing context also make simple “food appeared before symptom” counts misleading.
The product needed two things at once: a low-friction way to capture messy real life, and an evidence policy cautious enough to distinguish a possible association from a diagnosis or proven trigger.
What I built
The app, analysis layer and business system around it
Manual quick-add paths
Meals, symptoms, bowel movements and medications can still be recorded without AI, so the core tracker remains usable when a structured form is faster.
Evidence-aware insights
Numeric evidence is computed from the diary, while the narrative layer is instructed to surface possible associations, confounding and uncertainty rather than invent certainty.
Appointment reports
Users can turn their own records into a calmer summary for a clinician, with patterns and context organized without diagnostic claims.
Payments and entitlement
One-time purchases through Dodo Payments or Gumroad connect to email-based access, refund handling and customer lifecycle automation.
Privacy and control
The PWA uses passwordless access, zero-retention AI processing, data export and separate tracking-data or full-account deletion paths.
Product decisions
Building trust meant designing the boundaries too
I separated the public marketing site from the application runtime, kept structured entries as the source of truth, and treated generated insights as derived data that can be invalidated and recomputed. The app runs as a SvelteKit service behind Nginx, with SQLite, PWA support, payment webhooks, transactional email and automated backups.
The harder product work is not adding more AI prose. It is deciding when the evidence is too weak to say much, preserving user edits, handling failed submissions honestly, and keeping access and refund state consistent across several services.
The analysis layer remains under active evaluation and is not medical advice. Its caution rules describe the product standard I am building toward, not a claim of clinically validated conclusions.
Current scope
A connected product, not a feature demo
Find My Triggers now spans a public education site, installable web app, account and onboarding flows, voice and manual logging, analysis, reports, payments, email operations, privacy controls and recovery tooling. It demonstrates the kind of work I enjoy most: taking a messy personal or business problem and building the smallest coherent system around it.
Have a product idea with several moving parts?
I can help turn the workflow into a dependable web tool.
Start with the problem, the people using it and the systems it needs to connect.