AI/ML Lead / Asso Architect

As an AI/ML Lead / Associate Architect, your main responsibilities will include: - Designing, implementing, and optimizing end-to-end ML training workflows, from infrastructure setup to deployment and monitoring. This involves evaluating and integrating multi-cloud and single-cloud training options across platforms like AWS, as well as leading cluster configuration, orchestration design, environment customization, and scaling strategies. - Comparing and recommending hardware options such as GPUs, TPUs, and accelerators based on factors like performance, cost, and availability. - Having at least 4-5 years of experience in AI/ML infrastructure and large-scale training environments, being an expert in AWS cloud services, and having familiarity with Azure, GCP, and hybrid/multi-cloud setups. - Demonstrating strong knowledge of AI/ML training frameworks like PyTorch, TensorFlow, Hugging Face, DeepSpeed, Megatron, and Ray, as well as experience with cluster orchestration tools like Kubernetes, Slurm, Ray, SageMaker, and Kubeflow. - Possessing a deep understanding of hardware architectures for AI workloads, including NVIDIA, AMD, Intel Habana, and TPU. - Utilizing expert knowledge of inference optimization techniques such as speculative decoding, KV cache optimization, dynamic batching, quantization methods, model parallelism strategies, and inference frameworks like vLLM, TensorRT-LLM, DeepSpeed-Inference, and TGI. - Having hands-on experience with serving frameworks like Triton Inference Server, KServe, Ray Serve, and kernel optimization libraries like FlashAttention and xFormers. - Demonstrating the ability to optimize inference metrics, resolve GPU memory bottlenecks, and implement hardware-specific optimizations for modern GPU architectures like A100 and H100. - Leading the fine-tuning process of LLMs, which includes model selection, dataset preparation, tokenization, and evaluation with baseline metrics. This involves configuring and executing fine-tuning experiments on large-scale compute setups, documenting outcomes, and benchmarking against baseline models. If you have only worked on POCs and not production-ready ML models that scale, please refrain from applying for this position.,