Generative AI Engineer: Workflows & Master Prompting for Promotional Material

We are seeking a highly technical Prompt Engineer and Generative AI specialist to design the "backend brains" of an AI creative platform. We are not looking for someone to create one-off artistic images. We need an architect who can build "Master Prompts" and Custom Workflows that dynamically adapt to different inputs to produce high-quality promotional material at scale. You will be responsible for engineering the generation logic for two distinct generation pipelines: Text-to-Image (Concept-Based): - Create master prompt structures that accept dynamic variables (e.g., brand vibe, color palette, target mood, specific features) and output consistent, high-quality promotional imagery. - The system must reliably handle a wide variety of aesthetic styles—from corporate/sleek to playful/creative—without needing manual tweaking for every request. Image-to-Image (Reference-Based): - Design workflows that take a specific input image (a product shot, UI screen, or visual reference) and generate a promotional scene around it. - Critical Requirement: The workflow must maintain the integrity and recognizability of the reference image while seamlessly blending it into a generated background or context. We need you to craft complex, robust prompt templates (Master Prompts) that act as a wrapper around variable user data. You will determine: - How to structure the system instructions to ensure consistency. - How to handle negative prompting to prevent common failures. - How to inject "style" variables programmatically so the output matches the requested brand identity every time. Our priority is high-quality output using Open Source models. We are open to whichever open-source models you believe yield the best commercial results for promotional material. You should be expert in setting up workflows (e.g., pipelines, ComfyUI, or custom chains) that leverage these models effectively. If a workflow requires a closed-source model due to specific capabilities, we are limited to the following models due to per image API pricing: - Seedream 4.5 / 4.0 - Nano Banana (Gemini 2.5 Flash Image) - Vidu Q2 - Wan 2.5 Preview - FLUX.2 [pro] - Imagen 4 Preview 0606 If you prefer a model not listed here that is not more expensive than those listed here, we are open to discussion. Who We Are Looking For: - Programmatic Thinker: You understand that we are building a system, not just an image. You know how to write prompts that work as templates for variable inputs. - Workflow Expert: You go beyond simple prompting. You understand how to chain steps, use reference adapters, or fine-tune parameters to get the best result from open-source architectures. - Style Adaptability: You can create prompt structures that work equally well for a minimalist tech brand and a vibrant lifestyle brand. To Apply Please answer the following: - Describe your experience with creating "templated" or programmatic prompts that accept variable inputs. - Which Open Source models or workflows do you prefer for generating consistent promotional assets, and why? - Do you have experience with Image-to-Image workflows where preserving the reference image is critical? Please explain your approach. Apply tot his job