← Multimodal Sources 🕐 2 min read
Multimodal Sources

Making Magic With Gen AI: Capital One’s Prem Natarajan

> We rebuilt this fraud platform from the ground up and basically to use ML at the center of that enterprise and also to make it efficient so that we can make complex real-time ML decisions.

Show: Me, Myself, and AI · Publisher: MIT Sloan Management Review + BCG · Host: Sam Ransbotham, Shervin Khodabandeh

Episode URL: https://mitsmr.com/3RgPDrb

Publish date: 2024-01-03
Duration: NAs
Default source credibility: HIGH — MIT SMR + BCG joint production. Named F500 CxOs on-record about production AI deployments. Academic/consulting co-brand keeps claims disciplined. Host has light BCG framing; guest metrics stay HIGH.

  • Capital One’s Prem Natarajan discusses the inflection point in AI with large language models and the importance of balancing product vision, science, and engineering.
  • Natarajan highlights the fraud platform at Capital One as a prime example of AI’s practical application, emphasizing real-time decision-making and continuous model updates.
  • The democratization of AI access across social spectrums is seen as a significant opportunity, with a cautionary note on the misconception of AGI being imminent.

Extracted quotes

# Credibility Speaker Org Timestamp Topic Quote
1 HIGH Prem Natarajan (Chief Scientist and Head of Enterprise AI) Capital One 13:05 02-corporate-tools We rebuilt this fraud platform from the ground up and basically to use ML at the center of that enterprise and also to make it efficient so that we can make complex real-time ML decisions.
2 HIGH Prem Natarajan (Chief Scientist and Head of Enterprise AI) Capital One 16:31 07-adoption-challenges When we first started, it used to take months to update models. We now do it in days.

Per-quote detail

1. Prem Natarajan — Capital One (13:05)

We rebuilt this fraud platform from the ground up and basically to use ML at the center of that enterprise and also to make it efficient so that we can make complex real-time ML decisions.

  • Credibility: HIGH — Named exec at identifiable org with specific production deployment detail.
  • Topic tag: 02-corporate-tools

2. Prem Natarajan — Capital One (16:31)

When we first started, it used to take months to update models. We now do it in days.

  • Stat: Model updates went from months to days.
  • Credibility: HIGH — Named exec at identifiable org with specific metric and timeline.
  • Topic tag: 07-adoption-challenges

Extracted 2026-04-15T00:00:30 via scripts/podcast_mine.py (MLX mlx-community/Qwen2.5-32B-Instruct-4bit).