Overview

We're Hiring: AI Engineer — Image Generation, Model Training & GPU Computing

Responsibilities

• Image generation and processing: Develop workflows for image generation, editing and understanding. Work with diffusion or other suitable generative architectures, image preprocessing, classification, segmentation and feature extraction.
• Model training and fine-tuning: Train and adapt vision and generative models using PyTorch. Apply full fine-tuning or parameter-efficient techniques such as LoRA. Configure training loops, optimizers, learning rates, checkpoints and validation.
• Dataset preparation: Build pipelines for image cleaning, deduplication, annotation and augmentation. Maintain appropriate training, validation and test splits and prevent data leakage.
• Reinforcement learning: Implement and evaluate RL experiments for suitable tasks. Define reward functions, run policy optimization, compare against baselines and assess training stability and actual improvements.
• GPU development: Run training and inference on NVIDIA GPUs. Manage CUDA-compatible environments, monitor GPU memory and utilization, troubleshoot out-of-memory errors and apply mixed precision, gradient accumulation and checkpointing where appropriate.
• Evaluation and optimization: Track experiments and model versions. Evaluate output quality, image consistency, failure cases, inference latency, throughput and GPU memory usage.
• Deployment and delivery: Package inference workloads using Docker and integrate them with backend APIs and job queues. Add tests, monitoring and versioned releases, and support reliable deployment.
• Recommendation systems and applied research. Build and evaluate personalized recommendation and ranking systems using embeddings, user preferences and interaction data. Apply content-based, collaborative filtering or hybrid approaches as appropriate. Review relevant research and open-source implementations, validate promising methods through experiments, and measure recommendation relevance and ranking quality.
• Applied research: Review relevant research and open-source implementations, reproduce promising methods and demonstrate their value through measured experiments.

Qualifications

• Location: Gurugram — Onsite
• Employment: Full-time
• Cyurae is building an AI-powered consumer product at the intersection of fashion and technology. We are looking for an AI Engineer with hands-on experience in image generation, image processing, model training, fine-tuning, reinforcement learning and GPU-based development.
• You should be able to prepare datasets, run training experiments, evaluate model quality and deliver reliable inference workflows. You will work closely with our CTO, AI engineers and backend Founding team to turn model capabilities into production features.
• 3+ years of professional AI, machine learning or deep learning development experience.
• Strong Python and PyTorch skills, with practical understanding of neural networks, loss functions, optimization and training workflows.
• Hands-on image generation and image-processing experience, using tools such as Hugging Face Diffusers, Transformers, PEFT, OpenCV or equivalent libraries.
• Demonstrable experience training or fine-tuning models on project-specific datasets, including LoRA or another parameter-efficient method.
• Practical experience with NVIDIA GPUs, including training, inference, memory management and performance troubleshooting.
• Hands-on reinforcement learning experience, demonstrated through at least one implemented project or substantial experiment involving reward design, policy optimization and evaluation.
• Experience preparing datasets, tracking experiments and validating results on held-out data.
• Working knowledge of Linux, Git, Docker, model inference integration and testing.
• Willingness to work full-time onsite in Gurugram. Candidates outside Gurugram must be willing to relocate.
• Multi-GPU or distributed training using DDP, FSDP, DeepSpeed or equivalent tools.
• QLoRA for suitable quantized models, vision-language models and multimodal embeddings.
• Preference optimization methods such as DPO.
• Controllable image generation, image consistency, visual similarity or recommendation systems.
• AWS GPU deployment and experiment tracking using MLflow or Weights & Biases.