ML Systems Engineer – Multimodal & Foundation Models

About This Role

This role involves building multimodal AI systems within a large-scale production ecosystem, combining vision, language, and structured signals to drive automation. The engineer will operate at the intersection of transformer-based model architecture, applied AI research, and production ML systems engineering.

Responsibilities

• Design, train, evaluate, and deploy multimodal transformer models (vision + language) in scalable production environments
• Architect and implement multimodal models in PyTorch
• Design and maintain robust ML data pipelines
• Optimize training and fine-tuning workflows (SFT, RL-style iteration)
• Build scalable inference pipelines for large models (100M+ parameters)
• Improve evaluation frameworks and model observability
• Translate research ideas into production-grade systems

Requirements

• Strong PyTorch model architecture experience
• Hands-on experience implementing or modifying transformers
• Understanding of attention mechanisms and modern training strategies
• Experience building ML training/evaluation pipelines
• Experience deploying and serving ML models in distributed/cloud environments
• Ability to debug ambiguous model failures in real systems
• Experience reading and implementing research papers
• Familiarity with supervised fine-tuning (SFT)
• Exposure to reinforcement learning approaches (RLHF or similar)
• Experience with multimodal models (vision + language preferred)

Nice to Have

• Experience serving large transformer models
• Familiarity with distributed inference frameworks
• GPU optimization and performance tuning
• Experience with multimodal foundation models (e.g., CLIP-style, VLMs)

Skills

* Required skills