💥 Blast - Deploy AI safely and reliably at scale

TL;DR: Blast provides a platform to help enterprises turn their generative AI prototypes into reliable apps that can be confidently deployed at scale.

—

Hi everyone, we're Daniel and Arnav, and we're building Blast.

Problem: The mass adoption of LLMs across Fortune 500s is blocked by safety and reliability concerns.

Every large enterprise is trying to deploy their LLM-based apps/agent POCs at scale.

Unfortunately, prohibited content, hallucinations, and other failures leave enterprises open to brand damage and legal liability. For example, Air Canada was forced to honor an out-of-policy bereavement discount that its chatbot had offered to a grieving passenger.

Legacy enterprises in particular lack the talent and the tooling to solve these issues.

Solution:

We’re helping large enterprises build more reliable LLM apps.

1. Evaluation platform: Our red teaming tools enable developers to probe their end-to-end system with multi-turn conversations aimed at uncovering content/policy violations.
2. Governance models: We help enterprises detect, fix, and log failures in production.

We started two months ago, and we are piloting with a Fortune 50 company.

The Team:

👋 Hey, it’s Daniel and Arnav!

We met ten years ago in middle school and have been close friends ever since. We went to high schools in different states but reconnected at Stanford, where we were freshman year roommates. Daniel previously worked on AI security research at Stanford. Arnav previously worked on autonomous driving research at NVIDIA and in macro at Bridgewater.

Our Asks:

• We’d love intros to anyone working in security or AI governance at large enterprises. Email us at founders@blastsec.com
• If you’re interested in adversarially testing your AI apps/agents or evaluating your guardrails, we’d love to talk! Please email us above!