Building trust in AI with small language models: A conversation with Namee Oberst

In this episode of Catalyst, Tammy is joined by Namee Oberst, co-founder of LLMWare. Namee shares her unconventional journey from corporate attorney to AI technologist and her vision for making AI practical, private and trustworthy for enterprises. Tammy and Namee discuss small language models, AI for productivity and why human verification is nonnegotiable.

Highlights from the episode include:

• Small language models offer better control. Unlike large models trained on the entire internet, small models can be fine-tuned for specific tasks, stacked together and controlled with precision. This reduces hallucinations and increases trust.
• Supercharging customer service agents. Tammy and Namee push back on the idea of ripping out entire contact centers and replacing them with bots. Both have yet to have a customer service chat experience they find satisfying. They argue instead for supercharging agents with summaries, suggested next steps and fast access to knowledge. Then agents can focus on talking to customers and handling edge cases.
• Workflow automation requires domain expertise. Building effective AI agents is deeply personal because every team has unique procedures. No-code tools let small teams create and share their own workflows rather than having IT impose generic solutions that don’t match how people actually work.

Check out the full episode to hear more about on-device AI for enterprises and Tammy’s domain expertise in Thanksgiving pie baking.

Enjoying Catalyst? Don’t forget to like and subscribe for new episodes every Tuesday.

Please note that the views expressed in this episode may not necessarily be those of NTT DATA.