AI Research Scientist Generative Models Lead

Overview

We are seeking an expert in generative AI to design and train advanced architectures such as diffusion models, GANs, and transformers for photorealistic urban and interior design imagery. This role focuses on building core models from scratch, optimizing large-scale training pipelines, and innovating features like spatial layout preservation and material fidelity. Strong proficiency in PyTorch/TensorFlow, distributed training, and evaluation metrics (FID, CLIP-score) is essential.

Your role

Advanced Generative Model R&D

• Research and prototype cutting-edge generative model architectures (diffusion models, GANs, transformers) to produce high-quality Urban, architectural and interior design images from text prompts.
• Design and train the core network from scratch, innovating techniques to achieve photorealism, style accuracy, and structural consistency specific to urban design, buildings, and rooms.
  
  

Training & Optimization

• Implement and fine-tune the model architecture and training algorithms (data augmentation, loss functions, etc.), iterating rapidly to improve output quality.
• Explore novel features such as spatial layout preservation, material texture fidelity, and structured outputs that can be used downstream.
• Evaluate model performance with appropriate metrics (FID, CLIP-score, precision/recall, custom architecture-specific scores).
  
  

Collaboration & Mentorship

• Work closely with the Data/ML Engineer to debug issues and optimize training pipelines.
• Provide guidance on ML best practices, experimental design, and analysis; mentor junior engineers or contributors on complex model behavior.
  
  

Innovation & Research

• Stay at the forefront of generative modeling research, propose new architectures or training approaches, and prototype novel applications for the design industry.
  
  

About You

• Education: PhD (or Masters with equivalent experience) in Computer Science, Machine Learning, or a related field, with a focus on generative models, computer vision, or computational graphics.
  
  

Technical Expertise

• Proven expertise in neural network architecture design and algorithm implementation, especially for image generation tasks.
• Proficiency in Python and deep learning frameworks (PyTorch/TensorFlow); comfortable with libraries for vision and image processing.
• Strong mathematical and statistical skills for modeling and performance analysis.
• Generative AI Projects: 2–4 years (or equivalent) working on AI models for images — e.g., developing GAN-based image generators or training diffusion models.
• Large-Scale Training: Hands-on experience with large-scale image datasets and multi-GPU training regimes (distributed training, optimization, model compression).
• Design/Visual Domain: Experience in visual or design-oriented AI projects, such as architectural layout generation, style transfer for art/design, or photorealistic rendering.
• Cross-Disciplinary Collaboration: Background working in interdisciplinary teams (designers, artists, or other domain experts), demonstrating the ability to translate creative requirements into model improvements.
  
  

Reward & benefits

Explore the rewards and benefits that help you thrive – at every stage of your life and your career.

This Includes

• Comprehensive life insurance coverage.
• Premium medical insurance for you and your dependents.
• Generous annual leave balance.
• Flexible and hybrid work solutions.
• Remote work opportunities outside of country.
• Company gratuity scheme.
• Discretionary bonus program.
• Relocation assistance.
• Employee Wellbeing Program: 24/7 access to specialists in finance, legal matters, family care, personal health, fitness, and nutrition.
  
  

Seize every opportunity to sharpen your skills, expand your expertise, and be recognized for the impact you make.

About AtkinsRéalis

Were AtkinsRéalis, a world-class engineering services and nuclear organization. We connect people, data and technology to transform the worldsinfrastructure and energy systems. Together, with our industry partners and clients, and our global team of consultants, designers, engineers and project managers, we can change the world. Were committed to leading our clients across our various end markets to engineer a better future for our planet and its people.

Find out more.

---