← Multimodal Sources 🕐 2 min read
Multimodal Sources

VentureBeat in Conversation: Credit Karma’s path to scalable AI

> We use BigQuery, we use Dataflow very heavily for us to be able to process all this data at scale.

Show: Beyond the Pilot · Publisher: VentureBeat · Host: Matt Marshall, Sam Witteveen

Episode URL: https://traffic.megaphone.fm/UTEAU7636656651.mp3?updated=1761671029

Publish date: 2025-09-17
Duration: NAs
Default source credibility: HIGH — Named F500 practitioners on-record with production metrics. VentureBeat editorial vetting. Treat vendor-sponsored segments as MEDIUM.

  • Credit Karma processes over 65 billion model predictions daily, leveraging Google’s infrastructure for data consistency and security.
  • Vishnu Ram emphasizes the importance of data consistency across environments and the challenges of scaling AI infrastructure.
  • Credit Karma uses OpenAI models and a multi-cloud approach, with a focus on integrating AI with existing features to enhance user experience.

Extracted quotes

# Credibility Speaker Org Timestamp Topic Quote
1 HIGH Vishnu Ram (VP of Engineering) Credit Karma 03:17 02-corporate-tools We use BigQuery, we use Dataflow very heavily for us to be able to process all this data at scale.
2 HIGH Vishnu Ram (VP of Engineering) Credit Karma 02:53 01-ai-native-landscape We are running something like 65 billion model predictions daily.
3 HIGH Vishnu Ram (VP of Engineering) Credit Karma 13:04 07-adoption-challenges We have a dedicated cluster that we have for accessing open AI models that we know that the data that is going into those clusters are only available for our purposes to help our members.

Per-quote detail

1. Vishnu Ram — Credit Karma (03:17)

We use BigQuery, we use Dataflow very heavily for us to be able to process all this data at scale.

  • Credibility: HIGH — Named exec with specific tool details.
  • Topic tag: 02-corporate-tools

2. Vishnu Ram — Credit Karma (02:53)

We are running something like 65 billion model predictions daily.

  • Stat: 65 billion model predictions daily
  • Credibility: HIGH — Named exec with specific metric and denominator.
  • Topic tag: 01-ai-native-landscape

3. Vishnu Ram — Credit Karma (13:04)

We have a dedicated cluster that we have for accessing open AI models that we know that the data that is going into those clusters are only available for our purposes to help our members.

  • Credibility: HIGH — Named exec with specific deployment detail.
  • Topic tag: 07-adoption-challenges

Extracted 2026-04-15T01:28:22 via scripts/podcast_mine.py (MLX mlx-community/Qwen2.5-32B-Instruct-4bit).