AI/ML Engineer, Generative Previsualisation

REFINERY MEDIA PTE. LTD.; Singapore; On-site; SGD 90,000 - 130,000; Full time

Job summary

Refinery Media PTE. LTD. is seeking a creative AI/ML engineer to turn scripts, briefs, moodboards, and production intent into actionable visual outputs for directors and teams. You will experiment with generative image and video models, design training datasets, and shape model workflows for practical product features.

You will work closely with the founding team, product engineering, and production users to build AI systems that support creative decision-making across projects.

Qualifications

• Minimum 3+ years in ML, applied AI, CV, or AI engineering.
• Strong Python skills and practical experience with image/video generation.
• Familiarity with fine-tuning, LoRA, adapters, embeddings, or conditioning workflows.
• Experience turning messy real-world data into training or evaluation datasets.
• Ability to prototype quickly while considering production constraints.
• Understanding of model limitations, hallucinations, latency, and cost.
• Strong problem-solving and clear communication, with interest in creative tools.

Responsibilities

• Build and evaluate AI workflows for script-to-scene and script-to-shot previsualisation.
• Work with generative image and video models, multimodal models, and LLMs.
• Design datasets from scripts, briefs, visual references, storyboards, shot lists, moodboards, and metadata.
• Experiment with fine-tuning, LoRA/adapters, prompting, retrieval, and conditioning workflows.
• Improve consistency across characters, locations, style, mood, framing, and continuity.
• Explore workflows for generating visual options useful to directors and teams.
• Evaluate outputs for usefulness, quality, latency, and cost.
• Collaborate with engineers to integrate AI workflows into YAM.

Skills

Strong Python

3+ years ML/AI

Image/video generation

Multimodal AI

Visual storytelling

Clear communication

Prototype to production

Tools

Stable Diffusion

Flux

ComfyUI

Runway

Kling

V

Sora-style workflows

Job description

The role

This role is focused on turning scripts, creative briefs, visual references, moodboards, characters, locations, and production intent into useful visual outputs for directors, producers, and creative teams. It is not a pure research role, and it is not a generic chatbot role. We need someone who can experiment with generative image/video models, structure training and evaluation data, improve visual consistency, and help turn model workflows into practical product features. You will work closely with the founding team, product engineering, and production users to build AI systems that support real creative decision-making.

What you'll do

• Build and evaluate AI workflows for script-to-scene and script-to-shot previsualisation
• Work with generative image and video models, multimodal models, and LLMs
• Design datasets from scripts, briefs, visual references, storyboards, shot lists, moodboards, frames, and production metadata
• Experiment with fine-tuning, LoRA/adapters, prompting, retrieval, and reference-conditioning workflows
• Improve consistency across characters, locations, style, mood, framing, and scene continuity
• Explore workflows for generating visual options that are useful to directors, producers, and creative teams
• Evaluate generated outputs for creative usefulness, consistency, quality, latency, and cost
• Collaborate with full-stack engineers to integrate AI workflows into YAM
• Help decide when to use existing models, fine-tuning, custom datasets, third-party APIs, or open-source model pipelines
• Build repeatable evaluation methods so we can improve model outputs over time

What we're looking for

• 3+ years of experience in machine learning, applied AI, computer vision, generative AI, or AI engineering
• Strong Python skills
• Experience with image generation, video generation, multimodal AI, or visual-language models
• Familiarity with model fine-tuning, LoRA, adapters, embeddings, or conditioning workflows
• Experience working with messy real-world data and turning it into usable training or evaluation datasets
• Ability to prototype quickly while thinking about production constraints
• Practical understanding of model limitations, hallucinations, consistency issues, latency, and cost
• Strong problem-solving ability and clear communication
• Interest in film, content production, creative tools, or visual storytelling

Nice to have

• Experience with Stable Diffusion, Flux, ComfyUI, Runway, Kling, V, Sora-style workflows, or similar generative tools
• Experience with ControlNet, IP-Adapter, reference conditioning, character consistency, style transfer, or visual continuity workflows
• Experience with image/video understanding, shot analysis, scene parsing, or visual metadata extraction
• Experience with LLMs, structured outputs, RAG, or agentic workflows
• Experience building AI products beyond notebooks and demos
• Experience in film, advertising, animation, storyboarding, previsualisation, creative tooling, or content production
• Experience deploying AI systems into production environments

You’ll thrive here if

• You are excited by the idea of AI helping creative teams visualise ideas earlier
• You care about whether generated outputs are actually useful, not just impressive
• You enjoy experimenting with models but can stay grounded in product needs
• You can work with ambiguity and turn rough creative workflows into structured systems
• You are pragmatic about using the best available tools instead of reinventing everything
• You want to work at the intersection of generative AI, storytelling, and production workflows