The problem: Authors need a consistent social presence, but producing images, copy, links, and scheduled posts can become a second job. I built StoryFlow to turn that recurring work into a structured, reviewable publishing workflow.

• Owned the product end to end: concept, workflow design, positioning, pricing, customer-facing implementation, onboarding, subscriptions, infrastructure, launch, and ongoing operations.
• Connected the entire workflow: Amazon author/book ingestion, structured context, strategy-driven copy and image generation, review and approval states, queues, and Facebook/Instagram publishing through n8n, APIs, OAuth, and webhooks.
• Built for production: validation, error handling, human-review gates, monitoring, backup/recovery, and regression testing, with customer friction and operational feedback informing improvements.

A product decision: When users requested direct LLM prompting, I investigated the underlying need and identified persistent brand context as the more scalable requirement for consistent, customer-specific generation.

Configured operating workflow: six drafts daily, a controlled two-post queue, and two approved posts per day. The product includes the controls and lifecycle around generation—not just the model call.

The product: A guided, tarot-inspired AI subscription experience combining conversation, speech input, narration, generated imagery, and downloadable PDF readings. I owned the interaction design, AI orchestration, implementation, and recent backend modernization.

• Orchestrated multiple capabilities: conversational workflows, speech-to-text, text-to-speech, image generation, and PDF outputs through Flask and asynchronous Celery/Redis processing.
• Protected session state: per-reading isolation, stale-request and result checks, state-reset safeguards, and rate-limit handling to keep asynchronous outputs aligned with the correct reading.
• Modernized and released deliberately: migrated AI workflows to the OpenAI Responses API and used Git-based change review, isolated development/production services, validation checkpoints, targeted deployments, backups, and rollback procedures.

A product-engineering decision: I moved image preparation into background tasks with caching and reuse, while separating backend readiness from frontend presentation. That allowed preparation to proceed asynchronously without disrupting the intended narration and image transitions.

What to look for in the demo: how conversation, narration, imagery, and the final document come together in a guided experience. The creative context makes the interaction visible; the underlying work is orchestration, state management, reliability, and product design.