Built an AI system to interpret body composition data and transform therapeutic fasting into measurable results

The Problem

Traditional therapeutic nutrition programmes — including fruit-based detox and metabolic reset protocols — are powerful in principle but fragile in practice. They are rarely personalised, poorly quantified, emotionally volatile, and vulnerable to misinterpretation. People follow them hoping for results, but without real-time feedback, they are essentially flying blind. Scale weight fluctuations trigger panic. Plateaus trigger doubt. Rapid loss feels like success — even when it is muscle, not fat.

There is no intelligent feedback loop. No dynamic adaptation. No data interpretation layer.

What I Built

I inserted AI into the process — not as a calorie counter or generic chatbot coach, but as a real-time body composition analyst, pattern recognition engine, protocol translator, behavioural stabiliser, and strategic decision-support system.

The foundation was the fruit-based therapeutic nutrition method developed by Dr. Lyudmila Emilova, a Bulgarian medical doctor known for structured fruit fasting and detoxification programmes. Her principles include fruit as primary healing food, digestive rest, metabolic reset, structured mono-food phases, and gradual reintroduction.

Instead of following these principles passively, I engineered a measurable, AI-supported system around them — creating a hybrid model: traditional therapeutic nutrition × AI analytics × quantified self-tracking.

Methodology

I tracked daily body composition data using a Tanita body composition monitor and personal measurements — weight, body fat %, muscle mass, visceral fat rating, water %, bone mass, BMR, and circumference measurements. This data was fed into ChatGPT on a daily basis during active protocol phases, and weekly for structured trend reviews.

The AI performed multiple roles simultaneously:

Data Analyst — interpreting Tanita readings, distinguishing water loss from fat reduction, flagging muscle loss risk, and interpreting visceral fat trends.

Pattern Recognition Engine — identifying trends over weeks, highlighting plateaus, distinguishing normal fluctuations from meaningful shifts.

Protocol Translator — converting Dr. Emilova's general therapeutic principles into structured daily plans, measurable targets, and clear implementation steps adapted to my personal physiological response.

Behavioural Stabiliser — replacing the emotional volatility of dieting with analytical stability. A 1kg increase no longer triggered panic — it was identified as water fluctuation. A plateau was normalised and strategically handled. Rapid weight loss was examined, not celebrated.

Strategic Advisor — helping decide when to adjust caloric intake, evaluating whether protocol intensity was sustainable, and assessing metabolic adaptation signs.

Report Generator — compiling structured progress summaries, transformation narratives, and a comprehensive end-of-programme evaluation with all data organised for future reference and pattern tracing.

The project ran across four phases: (1) preparation and baseline setting, (2) active fruit-based protocol with daily AI monitoring, (3) adjustment and optimisation with muscle preservation focus, and (4) structured reflection, evaluation, and documentation.

Results

The outcomes were both measurable and behavioural:

On the measurable side — body fat percentage reduced, muscle mass was preserved during caloric restriction, and visceral fat trends improved. The data revealed that rapid early weight loss was largely water-based, true fat reduction became visible over time, and that scale weight alone was deeply misleading without composition data.

On the behavioural side — the transformation was cognitive. I shifted from emotional reactivity to analytical evaluation. I stopped reacting to scale numbers and started optimising body composition. I developed structured tracking habits, data literacy in personal health, and a preservation-focused mindset that replaced the typical "lose weight at any cost" approach.

The end-of-programme report documented the full journey systematically, creating a data baseline for future iterations of the protocol — meaning every future cycle starts with intelligence, not guesswork.

Tools

— ChatGPT: Primary AI partner for data analysis, pattern recognition, protocol translation, behavioural support, and report generation — Tanita body composition monitor: Daily body composition data (body fat %, muscle mass, visceral fat, water %, bone mass, BMR) — Personal measurements and tracking logs — Spreadsheet for data structuring and trend analysis

Why This Is an Impact Project

This project demonstrates something larger than personal health improvement. It is a proof of concept for how AI can augment traditional healing systems — making them safer, more measurable, more adaptive, and psychologically sustainable.

Without AI, this level of continuous body composition interpretation, trend recognition, protocol adaptation, and objective data analysis would require medical supervision or professional data analysis tools. AI democratised that capability.

The model is replicable. Anyone following a structured nutrition programme — therapeutic fasting, metabolic reset, clinical detox — could use this same approach to transform a subjective health journey into a quantified, iterative optimisation process.

Instead of replacing traditional medicine, AI augmented it. Instead of replacing human intuition, AI stabilised it. Instead of blindly following a protocol, I co-engineered it.

Key Learnings

1. AI as metabolic co-pilot, not replacement — the most powerful use of AI in health is not generating meal plans, but interpreting complex physiological signals in real time and translating them into actionable decisions.

2. The biggest transformation was psychological — shifting from emotional reactivity to evidence-based evaluation changed my entire relationship with health tracking. This cognitive shift is arguably more valuable than the physical results.

3. Traditional methods + AI = safer innovation — Dr. Emilova's protocol is clinically grounded but designed for general application. Adding an AI interpretation layer made it personally adaptive and measurably safer (particularly around muscle preservation during restriction).

4. Data infrastructure matters for future cycles — the end-of-programme report is not just documentation; it is intelligence for the next iteration. Each cycle gets smarter because the data carries forward.

5. Daily AI interaction creates analytical discipline — the habit of feeding data to AI and receiving structured interpretation daily built a form of health data literacy that persists beyond the programme itself.