Aditi S.

About Aditi S.

Software Engineer | AI Enthusiast | Technology Explorer Versatile and…

Experience

1. Software Engineer

   Microsoft

   Sep 2022 — Present

   Developed an automated model onboarding pipeline in Azure DevOps, reducing model onboarding time in AzureML Studio by 67%- Contributed to the foundation model inferencing engine on top of vLLM, enhancing performance and scalability by adding multi-modal support for Phi-4 multi-modal models- Designed deployment templates and containerized solutions for models with inferencing techniques like NVIDIA NIMs & TensorRT- Worked on Quota management service for Model Resource Providers for automated resource allocation for gpt models- Enabled model onboarding on AI Studio using Bring Your Own Container (BYOC) for custom models- Benchmarked models such as Meta LLaMA-3, Phi-3.5, NVIDIA Triton, and AI for Science in the AzureML Model Catalog- Developed LLM-as-a-Judge metrics ( GPT-Relevance, Groundedness, Responsible AI) through prompt engineering techniques, optimizing evaluation for text generation and question-answering tasks- Contributed to the azureml-metrics library by implementing Python-based evaluation metrics- Designed and automated an end-to-end task-to-metrics mapping system for NLP & NLG tasks (e.g. RAG, chat completion, text generation, and translation), significantly reducing manual effort.

Education

• Georgia Institute of Technology

  Master of Technology - MTech, Artificial Intelligence

• National Institute of Technology Hamirpur-Alumni

  Bachelor of Technology - BTech, Computer Science

  2018 — 2022