Reliability is the next chapter of AI infrastructure.
obsrv is a product by thetalab. It exists to make the runtime side of AI as boring and trustworthy as the runtime side of any other production system.
thetalabis a small team of distributed-systems engineers and applied AI researchers who've operated production AI at scale — through outages, regressions, and the occasional 3am page about a hallucinated billing endpoint. obsrv is our first product.
Every engineer at thetalab has run on-call for an AI product. That shapes everything we build. Our SDKs fail soft because we know telemetry going down can't take production with it. Our storage paths are tenant-scoped because we know data export requests are real. Our cluster discovery is unsupervised because we know taxonomies don't survive contact with traffic.
Backed by
thetalab is backed by founders and operators of Anthropic, Stripe, Vercel, Datadog, and Sentry — the people who built the previous generation of developer infrastructure.
Where we are
We're a remote-first company with a hub in San Francisco. We hire engineers anywhere in the Americas and EMEA. If reliability is your favorite word, come build it with us.
Ship the agent.
obsrv will record everything.
Drop the SDK in three lines. The recorder lights up the moment traffic starts flowing.
tr_01HXR4Z9CK
support_agent · 4.2s
✗ wrong_order