Mobile AI Product
AI Engineering · Software Engineering
Overview
Full product build for a personalized AI image generation app: Django backend, distributed GPU training pipeline, React Native mobile app, and analytics. Users upload photos to fine-tune a Stable Diffusion model on their likeness, and the system delivers daily drops of stylized images across GCP and CoreWeave.
Challenge
The client had a custom Stable Diffusion model but no production infrastructure. Each user required a dedicated fine-tuned model (~2GB), training had to complete quickly after onboarding, and daily image generation needed to run reliably across thousands of users. The mobile app, backend API, GPU orchestration, and analytics all needed to be built from scratch.
Solution
We built the Django backend and distributed task system that orchestrates the full pipeline: user onboarding triggers model training on uploaded images, trained models are stored and managed per user, and a daily generation job produces personalized image drops across styles. GPU workloads are distributed via Celery and Redis across instances on GCP and CoreWeave. We led the React Native mobile engineering team, implemented push notification schemes for daily drops, built comprehensive user engagement analytics, and centralized observability across the distributed pipeline with Cloud Logging.
Results
20+
Art styles per daily drop
Daily
Automated generation per user
<10 min
Upload to trained model
2
Cloud GPU providers orchestrated