← back
My AI Thinks I'm Eating My Feelings (and Other Nutritional Insights) - Rami Alhamad
Takeaway
Real-time user feedback toasts beat formal evals for early consumer AI, and persistent per-user memory plus task chunking are what makes interactions feel fast and personal.
Summary
- Rami Alhamad (Alma) shares lessons from 8 months building an AI nutrition companion: simple text-like tracking, persistent user context, eventual restaurant/product recommendations.
- Co-developed the 'Alma Score' (0-100 holistic food quality metric) with Harvard's Dr. Eric Prim — 1000 calories can be eaten a million ways with very different quality.
- Engineering lessons: real-time 'how did Alma do' toast beats offline evals; chunk LLM tasks into modular steps streamed to iOS for perceived speed; persistent 'about you' memory updated on each interaction.
- User signal: when streaks accidentally broke, support channels blew up — confirming feature stickiness; multimodal (voice/photo/text) input was preferred over any single modality.
consumer-ainutritionmemory
Original description
Eating well shouldn't require an advanced degree or hours decoding nutrition labels. Alma leverages cutting-edge AI to turn complex nutritional science into straightforward, personalized advice. In this talk, we'll dive into how we're using large language models and real-world user data to reshape how people track, understand, and improve their diets. We'll share insights on building user-friendly AI experiences, practical lessons from Alma's journey, and how we're making nutrition advice smarter, simpler, and genuinely helpful—one meal at a time.