What tech stack does Safe Superintelligence use?

The one firmly confirmed element is compute: Google Cloud is SSI's primary infrastructure provider, supplying Google's TPUs for training its models. This was announced alongside Alphabet's investment, making Google both an equity backer and the critical-path cloud vendor.

Nvidia is also an investor, which keeps a GPU relationship in the picture, but reporting consistently identifies Google TPUs as the primary training substrate. As a frontier research lab, SSI's backend is a Python-centric deep-learning stack running large-scale distributed training — standard for labs of this type, though SSI does not publish specifics, so anything below the Google Cloud layer is detected/inferred rather than disclosed.

SSI has no public product, so there is effectively no consumer-facing frontend stack to detect beyond a minimal marketing site and an Ashby-hosted careers portal (jobs.ashbyhq.com) for recruiting — the one confirmed SaaS tool besides Google Cloud.

Because the company has no sales or marketing organization, there are no detectable GTM tools — no CRM, sales-engagement, or marketing-automation signals exist publicly. Its "data" layer is internal training infrastructure rather than analytics or a data warehouse. We deliberately list only what has a real public signal and avoid inventing vendors.

The integration surface is narrow and Google-centric. Anything that complements Google Cloud and TPU-based training — observability, security, data pipelines, ML tooling — has a plausible fit; anything premised on AWS or Azure faces an uphill, displacement-only pitch.

On the GTM side there is nothing to displace: with no CRM or sales stack, SSI is not a buyer for revenue tooling. The build-vs-buy posture skews heavily toward build for anything research-core, so the realistic openings are recruiting tools (they use Ashby), security/compliance, and infrastructure that makes a small team's compute go further — all decided by a handful of founder-level people.