BLOG
Engineering notes from the team building Cheela.
Deep dives, product updates, and tutorials. Written by the engineers shipping the gateway.
AllEngineeringEvalsProductCompany
FEATURED · ENGINEERING · 6 MIN
How we route 40M requests a day without a queue
The scheduler behind Cheela's runtime router, and why we avoided a message broker entirely.
Priya Raman · Jul 14, 2026
router.select(runtime, load)
→ p50 8ms dispatch
Evals · 4 min
Grading agent output without writing a single rubric
A look at the eval pipeline that runs on every execution by default.
Dev Nair · Jul 9, 2026
Product · 3 min
Why we built capabilities instead of another SDK
The reasoning behind treating model calls as typed capabilities, not raw completions.
Priya Raman · Jul 2, 2026
Engineering · 7 min
Zero-downtime runtime failover, end to end
What happens on the gateway in the 400ms after a provider starts returning 503s.
Marcus Webb · Jun 26, 2026
Company · 2 min
Cheela raises a seed round to build the AI-native stack
What the funding changes, and what it doesn't.
Priya Raman · Jun 18, 2026
Engineering · 5 min
Token accounting across five providers, one schema
Normalizing usage metering so your bill makes sense regardless of runtime.
Dev Nair · Jun 10, 2026
Evals · 6 min
Catching silent regressions before your users do
How the eval pipeline flags a model swap before it reaches production traffic.
Marcus Webb · Jun 2, 2026