Filipe Meirelles

Applies domain knowledge developed across 5 fitness roles held during 2025-2026 to validate FitnessTech workflows and translate practitioner needs into product requirements.

Built a production AI-powered SaaS spanning 4 core workflows: AI training-plan generation, guided workout execution, progress analytics, and Pix/Boleto payments. Implemented structured LLM-output validation with Zod and fail-safe handling for invalid model responses.

Completed a 6-month technical-service assignment diagnosing device issues, performing repairs, and explaining technical outcomes to customers.

Delivered English instruction during a 3-month appointment, adapting explanations to different learner needs and proficiency levels.

Built a repository of AI-driven automations using n8n and OpenRouter, implementing agentic RAG-like patterns for data retrieval. Designed a Telegram-to-LLM response flow and documented 3 planned business use cases: message classification, lead qualification, and financial tracking.

Engineered an automated data pipeline using Python, MediaPipe, and OpenCV to process body landmarks into structured posture metrics and technical reports. Added confidence scoring, capture protocols, referral alerts, and offline processing to reduce false positives and keep sensitive images off cloud services.

Developed a guided breathing application supporting 1-10 rounds and 30-60 breaths per round, with configurable breath-hold periods. Implemented a state machine for 5 exercise phases, synchronized timers, audio cues, mute controls, and Android packaging preparation.

Completed a Bachelor's Degree in Physical Education, combining formal training in human performance with self-directed study in AI, automation, and web development.