Hire Ajay B. - MLOps Engineer

Megatron-LM and NeMo Framework

* ML Infrastructure and DevOps AMD MI GPU/ROCm * Distributed training and inference Performance and validation engineer specializing in large-scale distributed GPU training and inference on AMD Instinct MI GPUs with the ROCm software stack. Experienced in diagnosing performance regressions, automating cluster-wide benchmarking across training and inference workloads, and enabling customers to deploy and validate multi-node GPU clusters at scale. Ryzen AI NPU * Automate distributed training jobs on AMD GPUs and collect metrics. * Built automation of ML model testing and infrastructure pipelines supporting cutting-edge AI models (CNNs, LLMs, BERT, Diffusion) on AMD Ryzen AI. * Automated the build and CI model evaluation lifecycle for ML environments * quantizers, optimizers, compilers, and containerized (Docker) ML environments using Kubernetes, leading to faster, more stable development and releases.

* Setup of continuous integration for ML accelerator firmware and drivers, including Vitis AI, ensuring rapid and reliable software delivery. * Architected and implemented a robust build pipeline using Docker, Jenkins, Make, and Bash scripts, streamlining the development process and reducing build times. * Developed, tested, and deployed an internal build/test tool using Python and Django, while swiftly addressing any issues to maintain tool reliability. * Scripted the automation of Docker image builds and container deployments for the build/test tool, including managing Dockerfiles for PetaLinux environments. * Implemented CI/CD processes using Jenkins, along with custom Python and Bash scripts, to automate builds, validate code changes, and deploy new artifacts. * Collaborated closely with development and QA teams to ensure the timely delivery of high-quality builds and releases, consistently meeting project deadlines. * Converted an existing make build system to Gradle, establishing a modern CI/CD pipeline to validate incoming changes, deploy artifacts, and run comprehensive tests. * Setup infrastructure monitoring using Prometheus and Grafana.