Alert on symptoms
Pages fire on what the customer feels, not on what a dashboard happens to measure.
Console styleThis is a sample — build yours free
kiran@toronto:~$whoami
# staff_site_reliability_engineer
Kiran Mehta
Staff Site Reliability Engineer
SRE who treats an outage as a design review that arrived late.
- open_to:
- Staff and principal SRE roles, Remote or Toronto
- location:
- Toronto, ON
About
kiran@toronto:~$cat about.md
Six years keeping payment infrastructure up, and a habit of writing the postmortem nobody dreads reading. Happiest on a team that wants fewer pages, not more dashboards.
Workplace
kiran@toronto:~$open workspace.jpg
By the numbers
kiran@toronto:~$metrics by_the_numbers
ValueMetricNote
99.99%Uptime heldPayments tier, two years
4 minMedian time to detectDown from 22 minutes
70%Fewer pagesAfter the alert rewrite
38PostmortemsWritten and shipped
Pages per week
kiran@toronto:~$plot pages_per_week.csv
# average pages per on-call week, payments tier
How I work
kiran@toronto:~$cat how_i_work.md
Blameless, but specific
The postmortem names the system that failed and the change that stops it, never the person on call.
Automate the second time
Does it by hand once, writes the runbook, then deletes the runbook by automating it.
Experience
kiran@toronto:~$git log --graph -- experience
2023 to nowStaff SRE(current, Kelsey Pay)
- Owns reliability for the payments tier
- Led the alerting rewrite that cut pages by seventy percent
2020 to 2023Senior SRE(Latchbox)
- Built the incident process and the on-call rotation from nothing
2018 to 2020Backend engineer(Vero Logistics)
- Rewrote the dispatch service that kept falling over at peak
On call
kiran@toronto:~$cat on_call.md
A page has to be urgent, actionable and real, or it gets deleted. Every alert has an owner and a runbook, and the runbooks are reviewed every quarter.
Hand-offs are written, not verbal. Whoever comes on reads one page and knows what is burning and what is only smoking.
Stack and tooling
kiran@toronto:~$ls stack_and_tooling/
- Kubernetes
- Terraform
- Go
- Prometheus
- Grafana
- OpenTelemetry
- PagerDuty
- PostgreSQL
Systems I have owned
kiran@toronto:~$ls systems_i_have_owned/
- Payments tier
- Dispatch service
- Global CDN edge
- On-call programme
What colleagues say
kiran@toronto:~$cat what_colleagues_say/*
Kiran rewrote our alerting and the on-call rotation stopped burning people out. Retention on that team went up.
His postmortems are the only ones I read end to end. They say what broke and what changes, and nothing else.
kiran@toronto:~$./contact.sh
Let's talk reliability
Postmortem samples and references on request.
kiran@toronto:~$
Like this page? Build your own at pitchpage.co