Developed backend infrastructure for AI-powered ad creative generation, integrating Google Veo 3.1 for video and Ideogram for images, using text prompts and input images through a Python service. -Integrated OpenAI to power budget allocation and campaign optimization, recommending advertising channels and audience targeting based on campaign and job requirements.
Used LangChain to structure and template prompts, keeping generation requests consistent across ad workflows.
Designed asynchronous AI generation workflows with job tracking, webhooks, media processing, SSE-based progress streaming, and WebSocket notifications, connecting AI services, backend jobs, and frontend events.
Fixed bugs and maintained data synchronization across the platform's ad management and recruitment modules.