AI Voice Calorie Tracker with Social Accountability Feed

The client came to Peppernode with a clear concept and a fixed budget: a calorie and activity tracking app built around a personality-driven AI voice bot, with a casual social layer for sharing progress with a small circle of friends. He had the vision mapped out — we provided the design, development, and technical judgment to bring it to life.

The client wanted a fitness app for people who don’t want to count macros — casual users who just want to stay roughly on track without it feeling like a chore. The concept combined a voice-first AI bot with a distinct sarcastic personality, calorie and workout logging, and a private social feed for a small group of friends. No existing product offered all three in one place.

The hardest problem was calorie accuracy. AI models returned inconsistent estimates for the same food across repeated queries, and off-the-shelf databases had gaps — especially for branded and processed foods. Getting the bot to log reliably while also sounding like itself was the central challenge the team had to solve.

• No existing fitness app matched this combination: AI voice logging, personality-driven bot, and casual social feed in one product
• Building a native iOS and Android app was out of budget at the start
• Off-the-shelf calorie databases were either too expensive, poorly organized, or not accurate enough for realistic food inputs
• AI models hallucinated calorie counts — the same meal could return wildly different numbers across repeated queries

The bot was the product. If it couldn’t log food reliably and respond in a consistent voice, the whole concept fell apart. A fun but inaccurate bot is entertaining once. After that, users stop trusting it — and stop using the app.

Getting calorie accuracy right was critical, but the client was equally clear that personality mattered just as much. A technically precise bot with a flat personality would miss the point entirely. The team had to solve both problems at the same time, on a timeline that didn’t allow for endless experimentation.

• 1 Project Manager / BA
• 1 UI/UX Designer
• 1 Backend Developer
• 1 Frontend / Mobile Developer
• 1 QA Engineer

The project launched as a Progressive Web App to move fast and keep costs down — building natively for iOS and Android from day one would have doubled the budget. As the product vision sharpened, the architecture evolved: the web app was wrapped inside a native mobile shell using a hybrid PWA/WebView approach, delivering a full App Store and Google Play presence without a full rebuild.

Throughout both phases, the team worked closely with the client, flagging trade-offs and adjusting priorities as the product developed.

• AI Voice Bot — Talk to Log Tap one button, speak naturally. The bot listens, logs your meal or workout, and talks back — with a sarcastic edge. If you claim you ran 20km in 3 minutes, it’ll call you out. Manual text entry exists, but voice is the product.
• Live Subtitles Users can tap to see a real-time text transcript of the bot’s response — for noisy environments or users with hearing limitations. Added ahead of App Store launch to meet California mobile accessibility standards.
• AI-Powered Calorie Estimation After testing 8 different configurations — including a custom USDA-sourced food database, third-party calorie APIs, and multiple AI models — the team identified the most consistent LLM for calorie estimation and tuned it specifically for this use case.
• Private Social Feed Users invite friends into a shared group. Everyone’s food and workout logs appear in a single timeline — visible to the group, private from the rest of the world. Comments and reactions keep it casual.
• Progress Tracking A dedicated tab shows weekly calorie intake vs. expenditure and estimated macronutrient breakdown — protein, fat, and carbs — based on logged entries.
• Freemium Access The app launches free. Subscription pricing is planned but not yet live.

The project ran on a tight feedback loop between the Peppernode team and the client. The client was hands-on — reviewing builds frequently, giving clear feedback, and making decisive calls on scope and direction. The PM/BA managed communications and translated product decisions into development priorities, keeping both phases of the build aligned with the client’s evolving vision.

• Full PWA MVP delivered in 2.5 months
• 8 configurations evaluated for calorie accuracy before reaching a production decision
• Hybrid native app build completed and prepared for App Store and Google Play submission
• Custom food database built and validated against ~5000 test food items

• A differentiated product in a crowded market — competing on experience, not data depth
• Hybrid PWA/native architecture delivered full App Store and Google Play presence at a fraction of native rebuild cost
• Voice-first logging with a distinct bot personality sets the product apart from traditional calorie trackers
• Accessibility compliance built in ahead of launch, reducing legal exposure for a California-based product