← All speakers

Sharmila Chokalingam

Conference affiliation: Microsoft · 2024

Sharmila Chokalingam worked in product marketing for Microsoft’s Azure AI in 2024, explaining how enterprises could choose models for different business tasks through Azure AI Model Catalog. Her focus was the connection between model capabilities and operational needs: a customer-care assistant, a document-summarization tool and a demand forecast require different kinds of models and different ways to assess results.

On this page

In her joint Azure AI session with product manager Shubhi, Sharmila asked why customers needed a broad model catalog, whether they used multiple models, and what happened to proprietary data supplied to those models. Shubhi demonstrated the platform’s deployment and application-building features; Sharmila presented customer applications that explained the business case for model choice by task.

Matching models to business work

Sharmila’s examples extended beyond conversational AI and distinguished established uses from applications customers were exploring:

  • EY: professional tasks and ongoing evaluation. She described EY’s use of Azure OpenAI models and its exploration of Llama for tasks such as documentation and summarization. She highlighted a lesson from EY’s experience: adoption requires continuous evaluation of AI performance and responsible-use practices after the initial deployment.
  • CMA CGM: customer care. The logistics company used Mistral for a chatbot supporting customer-care agents. Sharmila reported reduced response latency and improved customer satisfaction, alongside plans to extend language models into invoicing, document analysis and email writing.
  • Bridgestone: demand forecasting. Bridgestone used Nixtla’s TimeGEN to predict monthly demand across its product portfolio. The example showed why a model catalog could serve business needs beyond text generation: a specialized forecasting model addressed prediction accuracy and the complexity of forecasting pipelines.

What happens to customer data?

Sharmila made enterprise data privacy concrete by asking whether Microsoft or model providers used customer data to train models, then returning to the safeguards available in AI Studio and Model Catalog. Her questions connected model selection to the conditions under which a business could supply its own information. Together with the customer examples, they framed enterprise adoption around practical decisions: which model fits the task, how its performance will be evaluated over time, and how proprietary data will be handled.

Read the topics behind these talks

1 conference talk

Loading more talks and highlights…

References