AI Specialist

Описание

Job Description

MGID is a global advertising platform that helps publishers monetize their audiences and enables brands to promote their products and services across the open web effectively. Using AI-powered technology, we deliver high-quality native, display, and video ads in brand-safe environments, balancing user experience and performance.

Every month, MGID reaches over 1 billion unique users and delivers 185 billion ads in more than 70 languages. Trusted by brands like Renault, Domino’s, Nestlé, and IKEA, we support every stage of the marketing funnel – from planning campaigns to optimizing performance.

Founded in 2007, MGID has grown to operate from 18 offices worldwide, with a team of over 800 colleagues across North and South America, Europe, and Asia. Our ongoing investments in technology, talent, and strategic partnerships have fueled five consecutive years of double-digit growth. We continue to evolve and expand to meet the ever-changing needs of the digital advertising ecosystem.

This is a creative-technical hybrid role where the specialist must be able to produce both one-off generations and automated workflows for scalable generation of text, images, and video.

The key expectation is the ability to understand what creative outcome is needed, and then technically assemble the full process to achieve it: prompts, models, parameters, data structures, agent logic, and automation layers.

The role requires combining creative insight with technical execution to build generation workflows that reliably solve specific creative tasks and can be automated and scaled

Requirements

What You’ll Do

AI Pipeline Development & Orchestration

• Build AI workflows for generating text elements, images, and video based on input parameters and category requirements.
• Combine different models within a single process (LLM → image/video model → checks → additional processing steps).
• Configure agent-based scenarios, branching logic, variation patterns, and rule-driven flows.
• Transform individual generation tasks into stable, repeatable, and scalable processes.

Prompt Engineering & Model Integration

• Create system prompts, templates, and prompt chains that produce predictable and controlled results.
• Understand model limitations and how parameters, context, instructions, and API behavior affect output quality, variation, and task alignment.
• Select appropriate model settings for different generation types — textual, visual, and video.
• Configure rule-based, variable, and multimodel generation flows.

Data, Formats & Context Management

• Create and maintain clear input/output formats (JSON, schema-based prompts).
• Develop data structures for generation rules, campaign context, category information, and other supporting parameters.
• Use vector stores and other tools when necessary to support context-driven generation (RAG approaches)
• Ensure correct data transfer between models, agents, and automation layers.

Model Evaluation, Fine-Tuning & Quality Assessment

• Test and compare models for text, image, and multimodal tasks.
• Use fine-tuning and LoRA to adapt models to MGID’s specific needs.
• Define rules, prepare and structure dataset for fine-tuning and LoRA adaptation, including selecting appropriate examples, balancing categories, ensuring data quality and consistency, and formatting datasets according to model requirements.
• Evaluate output quality both technically and from the perspective of creative effectiveness and advertising requirements.
• Create and implement stability and quality metrics, and build iterative feedback loops that allow adjusting workflows and improving results based on testing data.

Automation & Workflow Optimization

• Analyze existing workflows and identify areas where automation provides significant value.
• Build utilities, micro-flows, and automation processes in n8n or other low-code/no-code tools.
• Support the deployment of new generation flows to production together with the dev team.

Documentation & Knowledge Transfer

• Create documentation for workflows, prompts, data structures, generation rules, and scenarios.
• Document processes and instructions, and share knowledge with the team.

Who You Are

• You combine creative understanding of the desired output with technical skills for building AI workflows.
• You understand the capabilities, strengths, limitations, and practical use cases of generative AI models (LLM, image, and video generation) and tools available on the market, and can select the right approach depending on the requirements of the task.
• You have experience working with LLMs and generative model APIs.
• You understand prompt logic, instruction structures, and data-driven generation.
• You think systematically and understand how model choices and parameters affect creative outcomes.
• You can assess the quality of visual and textual outputs with advertising requirements in mind.
• You communicate effectively with creative, product, and technical teams.
• You have basic Python knowledge or the ability to work with low-coding artifacts at the level of modifying existing fragments

Nice to have:

• Basic understanding of HTML and the ability to make edits when needed.
• Experience with SQL, data handling, and integrations.
• Experience with AI-IDE tools such as Cursor or Claude Code.
• Understanding of AdTech/MarTech and advertising format requirements.

MGID is a global advertising platform helping brands reach unique local audiences at scale. In MGID we empower brands and publishers to work together transparently through our privacy-first targeting technology to enable advertisers to drive performance and awareness, and publishers to retain and monetize their audiences. Today, we’re creating unique technologies and with your help, we are looking to aim even higher.