Machine Learning Street Talk (MLST) avatar

Dexa/Machine Learning Street Talk (MLST)

Learn more
Published Sep 12, 2024

Cohere's SVP Technology - Saurabh Baji

Cohere's SVP of Technology, Saurabh Baji, explores the company's pioneering strategies in deploying large language models for enterprises, stressing data security and enhanced performance through retrieval-augmented generation. He shares insights into effective model customization and optimization, illustrating practical implementations and cost efficiencies across diverse industries.
Episode Highlights
Machine Learning Street Talk (MLST) logo

Topics covered

Popular Clips

Questions from this episode

Episode Highlights

  • Efficient Customization

    Cohere's approach to model customization emphasizes efficiency and effectiveness, minimizing resources while achieving tailored results. explains that their models allow for full fine-tuning, enabling high performance at a lower cost by running multiple fine-tunes on a single GPU 1. This innovation allows customers to achieve their specific needs without the burden of large, impractical models.

    We have literally customers who are running 50 fine tunes on one single GPU. And so that lets you get away with as much customization as you want and yet keep the footprint as if you were just running a single one.

    ---

    Additionally, Cohere's tailored automation capabilities enhance user workflows by generating SQL, API calls, and more, making processes more efficient and customized to individual needs 2.

       

    Customer Collaboration

    Cohere places a strong emphasis on collaborating closely with customers to tailor models to specific needs and use cases. highlights the importance of separating core training workflows from customer-specific requirements, ensuring that feedback from customers informs model development 3. This approach allows Cohere to quickly address customer needs while also improving their core models.

    We learn from the field, we see what is potentially just a one off case. What are we seeing on the recurring basis? There's extremely valuable information because we are effectively seeing which way the market is going.

    ---

    The collaboration extends to various industries, with customers like Oracle and Salesforce utilizing Cohere's models for HR, sales, and finance applications 4.

       

    Case Studies

    Real-world examples demonstrate the success of Cohere's customized models across different industries. discusses how Fujitsu applies large language models to various enterprise customers globally, focusing on multilingual capabilities 5. Oracle uses Cohere's models in applications like Oracle Fusion and Oracle Netsuite, spanning domains such as HR, supply chain management, and finance.

    Oracle uses the coherent models to create functionality in several applications that are part of these suites. And so several use cases within each application that are part of these suites as well.

    ---

    Baji predicts that the next year and a half will see a significant increase in public usage of these models, highlighting the growing adoption of AI in enterprise applications 6.

Related Episodes