ApexFit
AI Personal Trainer
Founder & Product Engineer
A consumer AI product built at Unilise Technology, designed to provide personalized workout programming and coaching. Architected and built entirely from zero as a solo engineer, covering every layer from React Native mobile applications to backend services, AI agents, subscription infrastructure, analytics and production observability.
Scope of ownership
What was built
- End-to-end product engineering as sole developer
- OpenAI-powered AI Coach with tool-based state access
- Multi-stage reliability layer around LLM-generated outputs
- Cross-platform mobile with Apple Health and Health Connect integration
- App Store and Google Play subscription infrastructure
- Full observability: PostHog, OpenTelemetry, Elastic APM, Prometheus, Grafana
- TypeScript monorepo with Turborepo, OpenAPI SDKs, Docker, CI/CD
- Private beta with 30+ testers ahead of public launch
Engineering approach
Reliability around probabilistic AI outputs
Multi-stage reliability layer including schema and domain validation, normalization, matching, algorithmic fallback and model-assisted recovery to ensure consistent, usable workout programs from LLM generation.
State-aware AI interactions
AI Coach designed with tool-based access to application state, enabling contextual conversations that work with workout programs, equipment, schedules, user preferences and coaching data.
Cross-platform health data integration
Integration with Apple Health and Android Health Connect for activity and health data synchronization across platforms.
Subscription state management
App Store and Google Play subscription infrastructure with server-side IAP webhook processing, entitlement management and subscription-state synchronization.
Product observability
Comprehensive observability spanning PostHog product analytics, OpenTelemetry distributed tracing, Elastic APM, ELK log aggregation, and Prometheus with Grafana for infrastructure monitoring.
Selected stack
Current state
Currently validating the product with a private beta cohort of 30+ testers ahead of public launch. Structured as a TypeScript monorepo with Turborepo and pnpm, OpenAPI-generated SDKs, automated testing, Dockerized deployments and GitHub Actions CI/CD.