fitness-tech

Turn authorized motion references into a consistent AI digital-human sports coach, reusable action library, timed course, and voice-guided training experience.

Momo — our first attempt at a digital store manager, and the start of this journey. Now carried forward as Saros (in preparation).

Multi-agent fitness coaching without the framework bloat. Phase-based reasoning pipeline (Analyze → Think → Summarize → Decide) that actually improves training plans — not just chat.

AI-powered movement analysis platform using computer vision and deep learning to classify lifts, analyze biomechanics, generate coaching feedback, and create visual phase breakdowns.

A multi-agent system in Python (Anthropic SDK) that generates personalized workout routines and meal plans for a personal trainer — free rule engine + optional LLM layer, explicit-state orchestrator, Streamlit dashboard, and deterministic safety validator.

AI-powered pull-up counter using TensorFlow.js MoveNet for real-time rep tracking in the browser.

Computer vision squat form classifier using MediaPipe Pose Landmarker, scikit-learn, and Streamlit.

Modern and responsive fitness-tech landing page for PulseFit, built with HTML5, CSS3, and JavaScript (ES6), featuring interactive UI, responsive design, animations, and client-side form validation.

A week to week CrossFit programming tool that captures workouts as structured data and surfaces cycle-level balance and priority coverage at planning time. Solo product and engineering case study.

DIY Smart Treadmill Sensor using ESP32-C3 & TCRT5000 optical sensor. Broadcasts speed & distance via Bluetooth BLE RSC to Garmin, Zwift, and more. Features an embedded Wi-Fi Web Dashboard & 3D printable enclosures.

AI fitness coaching app with Next.js, FastAPI, LangGraph agents, PostgreSQL/pgvector RAG, Azure Container Apps deployment, and cost-safe Gemini BYOK fallback mode.

FormLens v1.0.0 🏋️‍♂️ is a mobile-first fitness technique coach built to help users scan gym machines, learn correct form, and train with clearer confidence through animated guidance, audio cues, haptics, and a real Node.js backend.

Research MVP for latent physiological state estimation and continuous adaptive training control using a hierarchical beta-VAE and HJB optimization.

RepGuard - wearable gym movement form corrector (ESP32 + gyroscope + on-device ML)

2D computer vision system for Romanian Deadlift technique analysis and repetition-level feedback.

IPF and Wilks Score Calculator for powerlifters.

EMG-powered workout analysis with real-time rep detection, automatic set flow, and activation analytics.

💪 Product Spec: GymLabb fitness coaching platform — 5 epics, user stories, tech stack

RepTrack AI is a computer-vision fitness analytics app that counts exercise reps from uploaded video and live camera input using pose estimation and generalized movement logic.