Launch is a digital development company within NTT Data, a top 10 global IT provider. Launch partners with industry leaders to transform emerging technology into scalable, real-world products, working with a wide range of companies such as Jeep and DTE Energy.

Beginning a fitness journey can be overwhelming. Many people are unsure how to perform exercises correctly, and without guidance, it is easy to build poor habits or lose motivation. Personal trainers offer valuable feedback, but they are not always accessible, especially for beginners or people on a tight budget.

Our Smart Personal Trainer is a mobile application that provides users with guided workout tracking and personalized feedback using only a phone and its camera. After the workout is complete, users receive a performance score calculated from a pose tracking model along with clear feedback on how to improve.

The app also includes a conversational AI trainer that users interact with through a chat window. This trainer reviews workout results and adjusts future routines by recommending changes to exercise selection, intensity, and overall workload, much like a real personal trainer.

The feedback is based on the user’s workout history, supporting beginners as they build consistency while helping more experienced users refine technique and progress safely.

Our application helps novice and experienced exercisers alike build muscle and improve their health safely at low cost.

Our software is built for the iOS and Android mobile platforms, utilizing the Unity game engine in C#. The back-end services are written in Python using a Flask API. Real-time pose estimation is performed using MediaPipe BlazePose, which tracks 33 key points on a user’s body. The system includes a Llama 3.1 model through Ollama to generate workout feedback.