01 FOODBUDDY

2023 • AI HEALTH TECH

FOODBUDDY

The AI accountability partner you can eat with. Snap-to-track with zero friction.

AI Vision Health Tech GCP Social

Use arrow keys or buttons below to navigate

02 The Accountability Gap

PART I: THE PROBLEM 02 / 09

The Accountability Gap

Tracking calories is boring; failing your diet is lonely. For calorie and macronutrient conscious mobile-first users, classical apps are mundane and unrewarding.

The "Data Entry" Churn

Classical apps treat food as spreadsheet data. This boring, mundane task causes users to stop recording meals within 2 weeks.

No Social Cost

Eating a donut in secret feels free. Eating it when you have to notify a buddy creates positive friction.

03 Accountability as a Service

PART II: THE SOLUTION 03 / 09

Accountability as a Service

Elevator Pitch:
A social-first food tracker that uses AI Vision to lower friction and "Peer Pressure" to increase adherence.

Approach	Pros	Cons	Verdict
Voice Logging	Fastest input.	Volume/Portion estimation is wildly inaccurate.	Rejected
Barcode Scan	Perfect accuracy.	Useless for home-cooked meals.	Backup feature
AI Vision + Social	Zero friction + Social Motivation.	Latency constraints.	Selected

Social "Push"

One-to-many sharing.
Real-time notifications only on "Bad" choices.
"Better Judgement" recommendations.

04 Vision Pipeline

PART III: TECHNOLOGY 04 / 09

Vision Pipeline

Leveraging early access to GPT-4 Vision and Gemini Pro Vision.

Input: Raw image buffer from camera.
Processing: Cloud Function spins up on valid upload.
Output: Structured JSON with macros and confidence score.

const analyzeFood = async (imageBuffer) => {
  const prompt = "Identify food items, estimate mass, and calculate macros.";

  const result = await model.generateContent([
    prompt,
    { inlineData: { data: imageBuffer, mimeType: 'image/jpeg' } }
  ]);

  return parseJSON(result.response.text());
}

05 Frugal Architecture

PART III: TECHNOLOGY 05 / 09

Frugal Architecture

Hosted on a GCP e2-micro instance to keep costs near zero while maintaining scalability.

Component	Technology	Why?
Backend	Node.js / Express	Lightweight, huge ecosystem.
Database	Firestore	Flexible schema for varying food data structures.
AI Model	Gemini Pro Vision	Lower latency & cost tailored for multimodal.

06 Why It Was Archived

PART IV: RETROSPECTIVE 06 / 09

Why It Was Archived

The Unit Economics Problem

"While the 'snap-to-track' feature was magical, the cost of processing 3-5 high-res images per daily active user (DAU) made the specific free-tier model unsustainable in 2023."

However, the lessons learned in Context Engineering directly influenced later projects like FileOps.

07 Interface Preview

PART V: VISUALS 07 / 09

Interface Preview

🥑

(Demo video archived)

08 Technical Stack

APPENDIX 08 / 09

Technical Stack

Frontend

React Native

For cross-platform mobile reach.

Backend

GCP

Cloud Run & Cloud Functions.

AI

LangChain

Orchestration of vision & chat agents.

09 STATUS: UNMAINTAINED

STATUS: UNMAINTAINED

The code lives on in private repos, but the spirit of AI-assisted accountability continues.

Back to Playground

Foodbuddy

Project presentation