From: aidotengineer
Alma aims to simplify healthy eating by providing personalized nutrition through an AI-powered companion [00:00:10]. The company believes that traditional nutrition apps often require excessive user input for minimal output, making healthy eating challenging [00:00:17]. By leveraging AI, Alma seeks to solve this problem and create a truly unique nutrition companion [00:00:42].
Core Pillars of Alma’s AI Nutrition Companion
Alma’s vision for an AI nutrition companion is built on three core pillars:
- Simple Nutrition Tracking Alma strives to make nutrition tracking easy, natural, and conversational, similar to texting a friend [00:01:04]. This approach avoids the need for users to search through endless product lists or rely on unreliable photo recognition [00:01:13].
- Contextualized Guidance Once tracking is easy, Alma focuses on building strong, powerful context around the user, including their flavor profile, interests, habits, and hobbies [00:01:23]. This context helps steer users towards their health goals [00:01:33].
- Product and Meal Connections Starting in the second half of the year, Alma plans to connect users with specific products, restaurants, and meals that can help them achieve their nutrition goals [00:01:37].
Key Learnings and Product Features
Alma launched a closed beta less than two months after incorporation, gathering user feedback to shape its development [00:02:01]. Key insights and resulting features include:
User-Centric Development
- Reliance on User Feedback: While exploring evaluation metrics, Alma prioritizes real-time, in-the-moment feedback from users [00:03:01]. A “How did Alma do?” dropdown toast appears after every interaction to gather this feedback [00:03:17].
- Constraining LLMs: Large Language Models (LLMs) have high error rates when given open-ended tasks [00:03:39]. Alma addresses this by constraining LLMs to very specific tasks [00:03:48].
- Step-by-Step Processing: To improve user experience and perceived speed, Alma breaks down the information relay process into steps [00:04:06]. For example, when a user enters “banana,” Alma immediately recognizes it as a food item, extracts it, and sends it to the client, while other processes like matching with the USDA database run in the background [00:04:21].
- Valued Features: A recent incident where streaks were broken highlighted their unexpected importance to users, leading Alma to double down on such engagement features [00:05:03].
Enhanced User Experience
- Continuous Context Building: Alma constantly learns about the user by adding novel pieces of information from interactions to a personal knowledge dataset [00:05:37]. This prevents users from having to reiterate information, making the AI agent smarter over time [00:05:43].
- Proactive Insights: Alma proactively detects insights about users’ eating habits and surfaces information they might not know, such as food pairings that increase nutrient absorption [00:06:24].
- Multimodality: Users appreciate the flexibility to interact with Alma via voice, photo, or text depending on the context [00:07:07]. Providing multiple modalities is preferred over bullishly focusing on just one [00:07:42].
Alma Score
Recognizing that users desire a holistic understanding of their diet beyond just calories and macros, Alma developed the Alma score [00:02:14]. This score, developed with Harvard University’s Dr. Eric Prim, is a simple score out of 100 designed to guide users towards foods that are fundamentally good for their health and away from less beneficial options [00:02:31].
Future Focus Areas
Looking ahead, Alma is doubling down on three key learnings:
- Brand and Design: Code is becoming a commodity, emphasizing the importance of user-centric product design and visual appeal to stand out [00:07:59].
- Trust and Partnerships: Building user trust, especially as a new entity, involves identifying and partnering with trusted sources that have an aligned mission [00:08:26].
- Community and New Data: As the community grows, there’s interest in how others eat, fostering curiosity about food consumption habits [00:08:48]. Alma aims to leverage its users and members to create new, valuable, and interesting data, rather than relying on publicly scraped data that larger players likely already possess [00:09:30].