Undrstnd Labs: Building an Inference API After Y Combinator Said No

I co-founded Undrstnd Labs with Denis Cartin, starting as Co-Founder before taking on the CTO title as the company matured. We applied to Y Combinator — and got rejected. Rather than treat that as the end of the road, we launched the company independently in Rome, Italy instead.

We also hosted 2 events in Tunisia, giving free platform credits to participating teams.

The Product

Undrstnd Labs offers inference that's meaningfully faster and cheaper than OpenAI, running on bigger and better models — the core pitch being real cost and performance advantages, not just an alternative for its own sake.

• Fast, Cheap Inference: Built infrastructure to serve inference significantly cheaper than OpenAI, without giving up model quality.
• Bigger, Better Models: Offered access to larger, stronger models than typical cost-focused alternatives.
• Developer-First API: Designed the API around developer experience, with Vercel AI SDK compatibility so teams already building on that stack could adopt it with minimal friction.

From Rejection to Launch

Getting rejected from Y Combinator didn't end the company — it just meant building it a different way, self-funded and based in Rome, Italy instead of through the accelerator track.

---

Impact and Achievements

• Resilience: Turned a Y Combinator rejection into an independently launched, running company rather than shelving the idea.
• Product Positioning: Built a genuinely differentiated offering — faster and cheaper than OpenAI on stronger models, not a like-for-like clone.
• Ecosystem Fit: Made the API a drop-in option for the existing Vercel AI SDK developer base rather than requiring a bespoke integration.
• Community Reach: Hosted 2 events in Tunisia, giving free platform credits to participating teams.