The Problem During my own weight-loss journey, I realised that tracking calories consistently was harder than sticking to the diet itself. Most calorie tracking apps offered a good experience only after purchasing a premium subscription, while Indian homemade meals like dal, poha, sabzi, besan chilla or rajma either had inaccurate nutrition information or were missing altogether. Logging meals manually also became repetitive, making it easy to lose the habit after a few days. This led me to the following problem statement: "How might we help Indian users track their nutrition accurately without making the experience expensive or time-consuming?" Target Users NutriTrack is designed for people who want to build healthier eating habits without paying for premium calorie tracking apps. Primary users include: • People trying to lose or maintain weight • Gym-goers tracking calories and protein intake • Users who primarily eat Indian home-cooked meals • Beginners who need guidance rather than just calorie numbers User Stories The product was designed around a few core user stories. • As a user, I want to log my meals in less than 30 seconds. • As an Indian user, I want accurate estimates for homemade food. • As someone tracking fitness, I want to monitor calories and macros together. • As a beginner, I want suggestions on what to eat instead of only seeing remaining calories. • As someone sharing a device with family members, I want separate food histories without creating multiple accounts. Product Vision NutriTrack is an AI-powered calorie tracker that combines verified nutrition databases with AI to reduce the effort required to consistently track food. Instead of only helping users record calories, the product also guides them towards better food choices through personalised meal recommendations. Features: AI Food Recognition Users can upload a meal photo, and AI detects multiple food items, estimates portion sizes and calculates calories and macronutrients. Each detected item can be edited, removed or resized before logging. Why? Most meals contain multiple dishes rather than a single food item. Detecting each component separately significantly reduces manual effort while improving accuracy. Hybrid Food Search Every food search first queries Open Food Facts for verified nutritional information. If no reliable result exists, particularly for homemade Indian meals, the app falls back to Groq AI to generate nutrition estimates. Results are clearly labelled as either Verified or AI Estimate, giving users confidence in the data source. Personalised Nutrition Goals Rather than asking users to manually select calorie targets, NutriTrack calculates personalised calorie and macronutrient goals using the Mifflin–St Jeor Equation. Basal Metabolic Rate (BMR) For men BMR = (10 × Weight) + (6.25 × Height) − (5 × Age) + 5 For women BMR = (10 × Weight) + (6.25 × Height) − (5 × Age) − 161 Daily calorie expenditure is estimated using TDEE = BMR × 1.55 Goal-based calorie targets are then generated as: Lose Weight → TDEE − 500 kcal Maintain Weight → TDEE Gain Muscle → TDEE + 300 kcal Macronutrient targets are calculated using: Protein: 1.4–1.8 g/kg body weight Fat: ~27% of daily calories Remaining calories allocated to carbohydrates This removes guesswork and provides every user with nutrition goals tailored to their body and fitness objective. AI Meal Recommendations Most calorie trackers stop after telling users how many calories remain. NutriTrack instead analyses remaining calories, remaining macros and meals already consumed to recommend foods that best fit the user's daily goals. This shifts the experience from passive tracking to active nutritional guidance. Food History A calendar-based history allows users to review previous days, encouraging consistency rather than focusing on a single day's progress. Success metrics I'm building toward Log a meal in under 30 seconds AI search accuracy above 90% Day 7 retention above 30% Food detection precision above 85% Scope of the MVP To keep the MVP focused on validating the core problem, I intentionally excluded: Barcode scanning Cloud synchronisation Wearable integrations Social challenges Recipe generation Premium subscriptions These features were deferred because they did not directly contribute to reducing the effort required to log meals. Constraints that shaped these calls No backend: shippable solo in days, at the cost of cross-device sync. Browser storage only: zero infra cost, but data doesn't survive a cleared cache. Limited AI training data on Indian regional dishes: real coverage, variable accuracy, hence the Verified/AI Estimate split. Built solo, limited time: forced ruthless scope cuts.