Amisha Singhal

Enhanced performance and optimized resource utilization by implementing Omnia 1.7 and fine-tuning Lora and SFT models on NVIDIA H100 GPUs. Scaled Llama 3 (8B/70B) models across multi nodes environments to improve computational efficiency and model accuracy and performed pretraining on the same cluster. Orchestrated NVIDIA Base Command Manager and Kubernetes deployments, reducing cluster provisioning time by 60% and accelerating workload management efficiency by 45%, resulting in improved overall system performance. Deployed and managed microservices using Helm for efficient, repeatable application packaging, reducing deployment complexity and ensuring consistency across environments. Architected and implemented automated CI/CD pipelines using Docker, Kubernetes, Mini-Kube, and Terraform, enhancing infrastructure management and application deployment. Boosted scalability and accelerated deployment velocity of 40%, enabling faster, more reliable software delivery across environments. Containerized applications with Docker and orchestrated deployment on AWS ECS. Implemented CI/CD pipelines using Jenkins, Helm, and AWS EKS, enabling rapid, reliable, and automated application delivery across development, staging, and production environments. Developed Python-based automation tools for log aggregation, performance monitoring, and alerting, improving system observability and enabling faster response to issues. Optimized AWS infrastructure using Terraform for provisioning and scaling EKS clusters, ensuring cost-effective resource management and performance optimization.

Built ML-driven dashboards using Splunk Enterprise and Machine Learning Toolkit (MLTK), reducing data analysis time by 35%. Implemented smart clustering, forecasting, and anomaly detection models to improve operational insights.

Designed and implemented Python and Bash automation scripts, enhancing operational efficiency and reducing manual effort. Assisted in Kubernetes cluster management, ensuring high availability and efficient resource utilization.