About
Most of an incident is triage, not repair. We built Ember to change that ratio.
When the pager fires, the on-call should be deciding and acting, not reading six tabs of logs trying to figure out what changed.
Built by engineers who carried the pager
Elliot Reyes and Nadia Whitfield spent years on-call at large consumer platforms. They watched the same pattern repeat across every incident: the alert fires, a half-awake engineer opens six tabs, and most of the next forty minutes goes to reading logs and chasing the deploy feed, not writing the fix.
Triage is not the hard part. It is just the slow part. A modern language model reading normalized telemetry under time pressure is faster and more thorough than a tired person searching Kibana at 2am. The on-call's judgment, context and authority are what matter. The reading does not need to be their job.
They left to build the tool they wanted on those nights. Ember was founded in March 2022. The team is small, working on AI full time, and the product is live.
The team behind Ember
Elliot Reyes
Co-founder and CEO
Elliot led incident and reliability engineering at a large consumer platform, where they built the on-call rotation, runbook program and post-incident review process from scratch. They spent six years carrying the pager and eventually running the team that responded when things broke at scale.
Nadia Whitfield
Co-founder and CTO
Nadia built observability infrastructure and ML platform tooling at a distributed systems company, working across tracing pipelines, log ingestion and the internal tooling engineers used to debug production. They joined Elliot to build the reasoning layer that was always missing from the incident stack.
A few facts
Incorporated
Delaware, United States
Headquarters
New York, NY
Market
Worldwide, US home market
Team size
5 to 25 people
Focus
Building AI full time
Address
115 Broadway, Suite 1502New York, NY 10006United StatesWhat we believe
Grounded, not guessing
Every ranked cause cites the telemetry it used. When the evidence is thin, Ember lowers the confidence and tells you what to check next. It never asserts a cause it cannot support.
The human runs production
Ember drafts the fix and shows the exact commands. Nothing touches production without a person explicitly approving it. That is not a limitation, it is the design.
Model-agnostic by principle
No single vendor is right for every task or will be the best option in a year. A thin routing layer keeps the product independent and lets Scale teams bring their own endpoints.
Shorter incidents over shinier dashboards
Another pane in the NOC is not the answer. Ember fits where the incident already lives, does the reading, and gets out of the way so the on-call can decide.
Want to help build Ember?
We are a small team hiring engineers, researchers and designers who care deeply about incident response. Remote-friendly, US-based.