Transformer Model Builder AI/ML

Title: Transformer Model Builder AI/ML

Looking for an experienced AI/ML Transformer Model Builder to design, finetune, optimize, and deploy transformerbased models (LLMs and related architectures) for production AI services, with strong skills in training frameworks, model optimization, and MLOps.

Responsibilities

• Design and implement training workflows for transformer models including tokenization, dataset pipelines, and augmentation.
• Finetune and optimize models using techniques like LoRA, quantization, pruning, and distillation to meet latency and cost targets.
• Scale training with distributed runtimes and optimizers (e.g., DeepSpeed, ZeRO) and manage mixedprecision training for efficiency.
• Integrate models into production: build inference services, optimize serving (batching, caching, Triton), and implement CI/CD for model artifacts.
• Collaborate with ML engineers and product teams to define evaluation metrics, A/B tests, and monitoring for model drift and performance.
• Document experiments and maintain model registry with reproducible training configs and checkpoints.

Required qualifications

• 5+ years ML engineering experience with 3+ years working on transformer models and LLMs.
• Handson experience with Hugging Face Transformers and model hubs, tokenizers, and finetuning workflows.
• Experience with distributed training tools (DeepSpeed, Accelerate) and optimization techniques for large models.
• Strong software engineering skills in Python, PyTorch/TF, and production deployment (Docker, Kubernetes).
• Proven track record of shipping models to production with observability and rollback plans.