Senior AI ML Architect

About the Role

We are looking for a Senior AI/ML Architect based in Hyderabad with more than 10 years of technology experience. This role is a hybrid position requiring three days a week onsite presence. The successful candidate will lead the design and implementation of advanced AI and machine learning solutions, focusing on training and fine-tuning large language models (LLMs) and other deep learning models using enterprise-specific datasets. You will also develop scalable infrastructure for model hosting, serving, and lifecycle management on cloud platforms.

Responsibilities

• Design and implement AI/ML solutions for training and fine-tuning large models with domain-specific data.
• Assess and select open-weight models considering domain fit, capabilities, infrastructure needs, costs, and licensing.
• Create end-to-end pipelines covering data preparation, model training, evaluation, packaging, deployment, and ongoing monitoring.
• Hands-on fine-tuning using LoRA and QLoRA techniques, optimizing hyperparameters such as rank, learning rate, batch size, and quantization.
• Prepare and refine domain datasets through cleaning, formatting, synthetic data generation, deduplication, and quality checks.
• Utilize PyTorch, Hugging Face Transformers, PEFT, TRL, Accelerate frameworks for model training and tuning.
• Design and manage GPU-accelerated cloud training environments on AWS, Azure, or Google Cloud Platform, optimizing GPUs such as A100, H100, or L40S.
• Deploy and host models with production-grade inference frameworks like vLLM, SGLang, TensorRT-LLM, or TGI, ensuring high availability and cost-efficiency.
• Oversee the full model lifecycle including versioning, registry management, deployment, rollback, monitoring, and retirement.
• Implement MLOps pipelines through MLflow, Weights & Biases, Kubernetes, Docker, Ray, and CI/CD tools to automate training, deployment, and retraining.
• Monitor model and infrastructure metrics including performance, latency, throughput, GPU and memory utilization, and operational costs.
• Develop regression and evaluation frameworks to validate improvements over baseline models.
• Resolve issues related to model training, GPU resource management, deployment, and production environment.
• Collaborate with cross-functional teams including data scientists, ML engineers, data engineers, and cloud engineers to ensure smooth operationalization of AI/ML models.
• Establish reusable architecture patterns and engineering standards for model training and hosting tailored to domain requirements.
• Provide technical guidance and mentorship to engineering teams engaged in ML/LLM model training and deployment.

Required Expertise

• Extensive hands-on experience with PyTorch and Hugging Face Transformers for model training and fine-tuning.
• Proven background in training large language models and deep learning models using enterprise domain-specific datasets.
• Advanced knowledge of LoRA and QLoRA fine-tuning techniques.
• Experience with frameworks such as PEFT, SFT, TRL, and Accelerate.
• Familiarity with open-weight models including Llama, Qwen, Mistral, Mixtral, Gemma, DeepSeek, and Phi.
• Strong understanding of GPU utilization, training hyperparameters, dataset preparation, and model evaluation.
• Expertise in deploying and hosting ML and LLM models on AWS, Azure, or Google Cloud, including GPU infrastructure planning and management.
• Proficiency with containerization and orchestration using Docker and Kubernetes.
• Experience operating model serving frameworks like vLLM, SGLang, TensorRT-LLM, and TGI for scalable production deployments.
• Comprehensive knowledge of MLOps processes including MLflow, Weights & Biases, CI/CD automation, model versioning, and monitoring.

Qualifications

A minimum of 10 years of extensive experience in technology, particularly in AI/ML engineering, deep learning, and production-level model development.

Work Mode and Location

This position is based in Hyderabad and follows a hybrid work model requiring three days per week at the office.

Eligibility

Open to candidates with any graduate degree.