Tag: #productivity
Author: Riaz Aslam
Date: August 2026
1. Vision & What the App Does
Many fitness enthusiasts struggle with two recurring problems every week. First, they are unsure whether they are performing exercises with correct form, which can lead to poor results or even injury. Second, gym attendance tracking is often manual and unreliable, with no real verification of physical presence.
FitIntelligence AI is a web-based application designed to automate both of these tasks. It acts as an AI-powered biomechanics coach while also providing a secure, automated gym check-in system.
The application uses computer vision in the browser to track body movement and calculate joint angles during exercises. This structured movement data is then sent to an AI model for analysis, which returns real-time feedback to improve form. At the same time, the app verifies gym attendance using biometric and location-based validation.
Key Features
- Real-Time Exercise Form Analysis
The app detects body joints using webcam input and calculates angles such as knee flexion, elbow position, and torso alignment during exercises like squats and pushups.
- AI-Powered Coaching with Amazon Bedrock
Structured joint data is sent to Amazon Bedrock (Nova models), which generates personalized feedback such as posture correction and movement suggestions.
- Secure Gym Check-In System
The app uses Amazon Rekognition to verify the user’s identity and AWS Lambda to validate their GPS location within a predefined gym radius.
- Automated Workout Tracking
Workout data such as repetitions, timestamps, and form scores are automatically stored in Amazon DynamoDB for later analysis.
2. How I Built It
The application was developed as a full-stack web app combining frontend computer vision with AWS cloud services.
Development Process
The frontend was built using React with TypeScript (Vite) to create a fast and responsive user interface. The webcam feed is processed using HTML5 Canvas, where body keypoints are tracked and rendered in real time.
A custom biomechanics engine calculates joint angles using vector mathematics. For example, angles between joints such as hip-knee-ankle are computed using trigonometric functions ( atan2 ) to evaluate movement accuracy.
AWS Integration
The application connects to AWS services through API calls:
- When a user performs an exercise, joint angle data is collected and sent to Amazon Bedrock, which returns coaching feedback.
- When the user initiates a check-in, a facial image is captured and verified using Amazon Rekognition.
- At the same time, AWS Lambda validates the user’s GPS location against a predefined geofence.
- Verified attendance and workout data are stored in Amazon DynamoDB.
Challenges & Solutions
One major challenge was maintaining smooth real-time rendering without affecting performance. This was solved by using requestAnimationFrame for Canvas rendering instead of relying on React state updates.
Another challenge was preventing false gym check-ins. This was addressed by combining two verification methods: facial recognition with high confidence and GPS-based geofencing, ensuring that check-ins are both authentic and location-verified.
3. AWS Services Used & Architecture Overview
The system follows an event-driven architecture where different user actions trigger AWS services.
Architecture Flow
- The web client (browser) captures webcam and GPS data
- Exercise data triggers Amazon Bedrock for feedback generation
- Check-in action triggers:
- Amazon Rekognition for face verification
- AWS Lambda for geofence validation
- Verified data is stored in Amazon DynamoDB
- Media and artifacts are stored in Amazon S3
AWS Services Breakdown
- Amazon Bedrock (Nova Models)
Processes structured joint data and returns AI-generated coaching feedback.
- Amazon Rekognition
Compares user face images for identity verification during check-in.
- AWS Lambda
Executes serverless logic to validate GPS location within a gym boundary.
- Amazon DynamoDB
Stores workout logs including repetitions, timestamps, and form scores.
- Amazon S3
Stores image snapshots and pose-related data.
- AWS Amplify / S3 Hosting
Used to deploy and host the frontend application.
4. What I Learned
This project provided practical experience in integrating AI with real-time web applications. I learned how to use Amazon Bedrock for structured input processing and how to design prompts that produce useful, actionable feedback.
I also gained insight into combining multiple AWS services into a single workflow. Using Rekognition with Lambda demonstrated how multi-factor verification can be implemented effectively in real-world applications.
Additionally, optimizing performance for browser-based computer vision taught me how to balance frontend processing with cloud-based intelligence.
5. App Link & Source Code
This project demonstrates how a simple but annoying real-world problem can be solved using modern web technologies and AWS cloud services. FitIntelligence AI reduces the need for manual tracking and improves workout quality through automated, intelligent feedback.
Loading article