AI/ML Engineer Needed for Image Scoring and Ranking Models

# AI/ML Engineer Needed for Image Scoring and Ranking Models We are looking for a smart, motivated AI/ML engineer to help us build models that can evaluate visual content, such as product images, YouTube thumbnails, ads, and other image-based creatives. We have labeled image data where each image is rated on a 0–9 scale. We may also have pairwise A/B comparison data, where labelers choose the stronger image between two options. The goal is to train a model that can learn from these labels and predict which images are likely to score higher, perform better, or be selected in comparison tasks. ## What You’ll Work On You may help with one or more of the following: * Train a classification, regression, or ranking model using 0–9 image scores * Train a binary comparison model using A/B image data * Explore reward modeling based on labeled image outcomes * Evaluate model performance using metrics such as accuracy, correlation with ratings, ranking quality, or pairwise prediction accuracy * Help design the data pipeline for image-label datasets * Experiment with vision models, CLIP-like embeddings, multimodal models, or other computer vision approaches * Potentially use the trained model as a reward model for improving or fine-tuning image generation outputs * If full RL fine-tuning is too complex, help improve the scoring model itself through better training, labeling, evaluation, and iteration ## Ideal Candidate We are open to junior or early-career candidates, but you should have strong general intelligence, strong computer science fundamentals, and a serious interest in AI/ML. You do not need to be a senior researcher, but you should be able to learn quickly, reason clearly, and work through complex ML problems. Strong candidates may have experience with: * Python * PyTorch, TensorFlow, or similar ML frameworks * Computer vision models * Classification, regression, or ranking models * Pairwise comparison modeling * Reward models or RLHF concepts * CLIP, vision transformers, multimodal models, or image embeddings * Data preprocessing and model evaluation ## Availability Availability for some overlap with U.S. Eastern Time business hours is preferred. You do not need to be based in the U.S., but we would prefer someone who can communicate regularly and join occasional calls during EST/ET-friendly hours. ## Nice to Have Experience with any of the following would be a plus: * Training models from labeled image data * Ranking systems * Image generation models * Fine-tuning ML models * RLHF, reward modeling, or reinforcement learning * Building practical ML pipelines from messy real-world data ## What We Care About Most We care more about your ability to think clearly and learn fast than your years of experience. This project may involve some advanced AI concepts, but we are happy to work with someone junior who is highly capable, curious, and willing to figure things out. You should be comfortable reading papers, testing ideas, debugging models, and explaining your reasoning clearly. ## First Milestone A good first milestone would likely be: 1. Review the available labeled image dataset 2. Recommend the best initial modeling approach 3. Build a simple baseline model 4. Evaluate how well the model predicts image scores or A/B outcomes 5. Suggest the next steps for improving performance ## To Apply Please include: * Your relevant ML or AI experience * Any projects involving image models, classification, ranking, scoring, or labeled visual data * The ML frameworks you are most comfortable with * A short explanation of how you would approach this problem * Links to GitHub, demos, papers, or past projects if available We are looking for someone who can start with a practical baseline, learn from the results, and help us iterate toward a strong image evaluation system.