Centcept

About

A small studio for
big AI bets.

Centcept Technologies was founded in 2024 by a group of engineers and designers who'd spent the last decade shipping production ML at hyperscalers and high-growth startups.

We started Centcept because most "AI projects" still die in demo. We wanted a studio with the seniority to make hard architectural calls — and the discipline to keep shipping after the launch press dies down.

(01) — Principles

How we operate.

Outcomes over output

We measure ourselves by what your business changed, not how many features shipped.

Senior or nothing

Every project is staffed with people who've shipped AI at scale. No bench warming.

Evals are the product

If you can't measure it, you can't trust it. We build the test bed first.

Boring infrastructure

We pick the dullest, most reliable stack that still gets the job done.

Weekly shipping

Demoable progress every week, in production within 6.

Own the moat

Your data, your prompts, your IP. Always.

(02) — Timeline

A short history.

  • 2024 Q1Centcept foundedThree founders, one rented loft, six pilots in flight.
  • 2024 Q3First agent platform liveNebula OS ships to production, handling 70% of L1 tickets.
  • 2024 Q410 client engagementsEmbedded squads across fintech, media, logistics.
  • 2025 Q140+ shippedOur practice doubles. We hire three more principals.
  • 2025 Q2Eval framework open-sourcedOur internal harness becomes a community standard.

(03) — Stack

Tools of the trade.

OpenAIAnthropicLangGraphLlamaIndexPineconeModalReplicateVercelSupabasePostgresPytorchTritonCloudflareTemporalBraintrustLiveKitDeepgramHugging Face

(04) — Press

"Centcept is what happens when senior infra engineers stop tolerating demo-ware."

The Verge

"A rare studio that treats AI like the operating constraint it actually is."

TechCrunch

"Quietly shipping the future of agent products."

Fast Company

Want to work together?

Get in touch →