The problem
On-call still runs on tribal knowledge
Incidents are expensive
Every minute spent working out what changed and where is a minute of customer impact. The first ten minutes of an incident are usually archaeology.
Runbooks go stale
The steps live in a wiki nobody trusts, or in the head of whoever was on call last time. The commands point at clusters that were renamed months ago.
Generic AI agents guess
Off-the-shelf AI SRE tools read raw output and summarize it. Ask twice, get two answers. They struggle most on the infrastructure that pages you most: the custom parts.
The approach
Deterministic where it matters. Intelligent where it helps.
Claude handles the judgment. Code handles anything that has to come out the same way twice.
Claude reads the alert, picks the runbook, runs the commands and adapts when the data surprises it. Parsing, time alignment, spike detection and classification are done by tested Go and Python code. Any number that reaches an incident thread comes from code, not from a model reading raw output. An engineer reviews the result before anyone acts on it.
More on the approach, and how it compares with off-the-shelf AI SRE tools →
Services
Three ways to work together
3 a.m. readiness assessment
A short, fixed-scope review of your on-call: alerts, runbooks and recent incidents. You get a written report that ranks what to automate first and what it's worth.
$7,500 fixed · about 2 weeks Details → Core offerOps automation build
Codified runbooks Claude can follow, deterministic parsers for your signals, and the guardrails around them, built in your environment and handed over with tests.
From $36,000 · then from $2,500/mo Details → SpecialistDatabase reliability advisory
Senior Vitess/MySQL help on retainer: upgrade and migration plans, control-plane automation, and a second pair of eyes when production is on fire.
From $3,000/mo · 10 hours Details →Track record
Built and proven on one of the largest Vitess fleets
HASHJOIN's approach comes from its principal's work as tech lead for database operations automation at Slack.
Work done by HASHJOIN's principal as an employee at Slack. Figures as of mid-2026. See the work →
Who you work with
A principal, not a pyramid
Vitaliy Mogilevskiy
Principal · Staff Platform Engineer
Ten years of database platform automation at InVisionApp, Auth0 and Slack, and an Oracle DBA before that. Most recently tech lead on Slack's OPS Framework and Sherlock. Named inventor on a patent for cloud database provisioning. Every HASHJOIN engagement is done by him directly.
Meet the principal at mogilevskiy.dev →Find out what your next 3 a.m. page costs
Tell us about your stack and your on-call. In 30 minutes you'll know whether automation would pay off, and where to start.